#data-science-and-ml

1 messages · Page 401 of 1

lapis sequoia
#

I am sorry. But there's no internet on my pc rn. Let me try to generate a hotspot

serene scaffold
#

did you try using the KFold generator?

>>> from sklearn.model_selection import KFold
>>> X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
>>> y = np.array([1, 2, 3, 4])
>>> kf = KFold(n_splits=2)
>>> print(kf)
KFold(n_splits=2, random_state=None, shuffle=False)
>>> for train_index, test_index in kf.split(X):
...     print("TRAIN:", train_index, "TEST:", test_index)
...     X_train, X_test = X[train_index], X[test_index]
...     y_train, y_test = y[train_index], y[test_index]
lapis sequoia
#

will this work for you?

unknown.png
serene scaffold
lapis sequoia
# 
from sklearn.utils import shuffle
col_num = X.shape[1]

new_Ind = []
cur_MaxScore = 0
col_Ind_Random = shuffle(range(0,col_num), random_state=1)

for cur_f in range(0, col_num):
    new_Ind.append(col_Ind_Random[cur_f])
    newData = X.iloc[:, new_Ind]
    clf=DecisionTreeClassifier(max_features="sqrt", max_depth=4, min_samples_split=22, min_samples_leaf=9, random_state=0)
    cur_Score = cross_val_score(clf, X, targets, cv=5).mean()
    if cur_Score < cur_MaxScore:
        new_Ind.remove(col_Ind_Random[cur_f])
    else:
        cur_MaxScore = cur_Score
        print("Score with " + str(len(new_Ind)) + " selected features: " + str(cur_Score))```
#

that would not show show you outputs

lapis sequoia
serene scaffold
#

you can copy and paste that as well. remember that you're more likely to get help when you make it as easy as possible for answerers, and text is easier to work with than screenshots/pictures.

anyway, try using the KFold generator to make the folds, and then evaluate during each iteration of the KFold generator.

serene scaffold
lapis sequoia
#

I studied k fold generator. I just forgot about it

#

Is it not possible to scale the whole dataset though?

#

sc.fit_transform(X)

serene scaffold
lapis sequoia
#

Then there would be no need to do the loop at all

serene scaffold
lapis sequoia
#

cross_val_score(clf, X, targets, cv=5).mean()

lapis sequoia
#

I just discovered feature scaling today

#

I believe it doesn't matter in decision trees though.

serene scaffold
# lapis sequoia no

fit and transform are two operations. fit adjusts the parameters of an sklearn object, and transform modifies data in some way. whereas predict passes X data through a model and returns the predictions.

#

so, fit_transform isn't going to be a method of the decision tree classifier, because it's a model.

lapis sequoia
#

This what I found on YouTube

Screenshot_20220508-054321110_1.jpg
#

Aren't both returning scaled data?

serene scaffold
lapis sequoia
#

There's no mention of classifier as of now in code

#

Just transformation

serene scaffold
#

but you have clf=DecisionTreeClassifier(max_features="sqrt", max_depth=4, min_samples_split=22, min_samples_leaf=9, random_state=0)

lapis sequoia
serene scaffold
lapis sequoia
#

Index(['age', 'anaemia', 'creatinine_phosphokinase', 'diabetes',
'ejection_fraction', 'high_blood_pressure', 'platelets',
'serum_creatinine', 'serum_sodium', 'sex', 'smoking', 'time']

serene scaffold
#

can you show me a few lines of the CSV as text?

lapis sequoia
#

75,0,582,0,20,1,265000,1.9,130,1,0,4,1
55,0,7861,0,38,0,263358.03,1.1,136,1,0,6,1
65,0,146,0,20,0,162000,1.3,129,1,1,7,1
50,1,111,0,20,0,210000,1.9,137,1,0,7,1

#

works?

serene scaffold
#

yes

#

where is it that you're doing feature selection?

lapis sequoia
serene scaffold
#

I don't get how it's doing feature selection

lapis sequoia
#

It's taking in a single feature at a time. Checking if it's increasing the accuracy of the model. If yes, we take it in. If no then we discard it.

serene scaffold
#

well, it doesn't look like your newData variable ever gets used.

lapis sequoia
#

Oh shoot

#

Looks like I fucked up the code

serene scaffold
#

now you know how I feel every day

lapis sequoia
#

Man! That's what happens when you try to reuse someone else's code 😭😭😭

serene scaffold
#

this just goes to show that anyone can betray you

#

anyone

#

you can never let your guard down

lapis sequoia
#

I thought I would just copy it from the professors notebook and modify it

#

Oh god

serene scaffold
#

professors can't write python code if their life depended on it.

lapis sequoia
#

Btw. I will modify it. Back to feature scaling. If I don't do train test split. And do fit_tranform(X) and pass that transformed X in cross_val. Will it be right?

# 
X_transformed=sc.fit_transform(X) 
cross_val_score(clf, X_transformed, targets, cv=5).mean()```
serene scaffold
#

only one way to find out 😄

lapis sequoia
lapis sequoia
serene scaffold
lapis sequoia
#

Worked

#

Good job kolv

serene scaffold
#

Stelercus. Ste-ler-cus

#

or is it stel-er-cus

#

idk

lapis sequoia
#

What's Is/eum

serene scaffold
#

"he/him" but in latin

lapis sequoia
#

Okay. So i corrected the code. It's still giving me lower accuracy than what I got by selecting all features.

serene scaffold
#

🔥 🔥 🔥 🔥

#

it might be that decision trees can learn to ignore features that aren't helpful.

#

(though it would likely take additional computation time to figure out which ones those are.)

lapis sequoia
#

tysm

serene scaffold
#

yw

lapis sequoia
misty flint
languid stratus
#

Anyone good with TFX?

lapis sequoia
#

I am getting an accuracy boost taking p=2 instead of p=3 in knn. In the minkowski distance. But what effect does increasing p would bring in generalisation of model. Does taking p with the highest validation score suffice?

#

The boost is 78.94 to 79.60

lofty basin
lofty basin
#

output=[]
for i in range(int(input(""))):
a,b,c,d=map(int,input("").split(" "))
if (a+b+c)<=d:
output.append(1)
elif (d*2)>=(a+b+c)>=d:
output.append(2)
else:
output.append(3)
for i in output:
print(i)

lofty basin
lapis sequoia
#

Can anyone help me on this

#

Anyone

worldly dawn
lapis sequoia
#

No

worldly dawn
lapis sequoia
#

Why?

#

It’s unlimited tries

#

Get it wrong doesn’t affect me@

worldly dawn
#

if it's not a test, then there is nothing wrong waiting a bit

lapis sequoia
#

Just help me it’s not a test

worldly dawn
lapis sequoia
#

Okay u will give me the answer in 1 hour?

worldly dawn
#

I will help you figure it out in 1h.
It's better to help you understand it than spoon feed you an answer that you don't understand

lapis sequoia
#

I’m not waiting 1 hour for that

#

I’ll ask in a different server thanks anyways

worldly dawn
#

!rule 8

arctic wedgeBOT
#
Rules

8. Do not help with ongoing exams. When helping with homework, help people learn how to do the assignment without doing it for them.

lapis sequoia
#

It’s not an exam

#

What makes u think that?

#

Like I said it’s unlimited tries

#

I just don’t have hours to wait@for 1 problem

#

It’s 12 o clock where I live

safe elk
zenith hawk
#

Ok very quick question. So i have like 10 graphs of same event but they differ slightly. I wanted to somehow use graph neural network to get approximation of fitting function, is it a right approach? Maybe some libs for that exist alr? Thanks.

#

Or it will be easier to get all functions from each graph and get median?

wooden sail
#

that depends on what knowledge you have about the graphs. if they are different because they are afflicted with 0-mean noise, you can simply average them. if the noise is impulsive, median filtering could work. if you already know a model for what the graph should look like but not the parameter models, you can optimize for parameters (e.g. with machine learning or classical gradient/newton methods)

#

if you know the type of noise but not the model, you could try and make a denoising network that takes in several noisy versions of a signal you made yourself synthetically, and give the network knowledge of the clean version, too

cyan sierra
#

Hello. I was wondering, do we typically only use cross_val_score on training data?

serene scaffold
cyan sierra
#

Yes, it is already splitting the data

unique flame
#

Is there someone here who is currently exploring Object detection in python?

misty flint
#

one of our profs "preferred" people to do their projects in matlab

#

everyone ended up choosing python instead

safe elk
#

Nice go python

woven coral
#

anyone knows how to fix this???

lapis sequoia
#

Is that a test??

lapis sequoia
#

Hi everyone, my pytorch ai wont work and i've narrowed its issue down to the fact that it doesnt respond to reward, i've attatched the py file of which i use to calculate its next move and how it uses reward for that

#

and im wondering what i can do to make it respond to reward and calculate a move based on its reward

bronze flume
#

does anyone used ACO,PSO algorithms for classification ?

odd meteor
echo summit
#

Guys, im new to data science and im currently out of ideas, so need help 😰 . I found a solution with iterrows, but the script will running for 2 weeks, so i need better solution

# DataFrame A
# ID | NAME |   CITY   |  STATE  |
# 1  | Jeff | New York |    ""   |
# 2  |Harold|  Dallas  |         |

#DataFrame B
# ID |   CITY   | STATE |
# 1  | New York |  NYC  |
# 2  |  Houston |  TX   |
# 3  |  Dallas  |  TX   |

I need to check for every row if CITY from DataFrame A is in CITY of DataFrame B if it matches then return value from STATE column in matched row and assign to the DataFrame A

so in the end it would be

# DataFrame A
# ID | NAME |   CITY   |  STATE  |
# 1  | Jeff | New York |   NYC   |
# 2  |Harold|  Dallas  |   TX    |

atm i have

dfA = dfA.assign(state=dfA["CITY"].isin(dfB["CITY"]).astype(str))

but it only returns boolean. How would i combine it to return those states?

agile cobalt
#

!d pandas.merge

arctic wedgeBOT
#
pandas.merge 

pandas.merge(left, right, how='inner', on=None, left_on=None, right_on=None, left_index=False, right_index=False, sort=False, suffixes=('_x', '_y'), copy=True, indicator=False, validate=None)```
Merge DataFrame or named Series objects with a database-style join.

A named Series object is treated as a DataFrame with a single named column...
odd meteor
# woven coral

Confirm if there's any word token in your corpus. If your corpus list isn't empty, then the problem is from your wordcloud code

agile cobalt
#

!d pandas.DataFrame.join

arctic wedgeBOT
#
pandas.DataFrame.join 

DataFrame.join(other, on=None, how='left', lsuffix='', rsuffix='', sort=False)```
Join columns of another DataFrame.

Join columns with other DataFrame either on index or on a key
column. Efficiently join multiple DataFrame objects by index at once by
passing a list.
echo summit
#

😮 i'll check it, brb

paper oak
#

I want to start learning AI ML, so from where should I start? Suggest some good resources to start with

lapis sequoia
paper oak
#

would be great , if u give some advice

lapis sequoia
#

have you seen the resources section ?

#

!resources

arctic wedgeBOT
#
Resources

The Resources page on our website contains a list of hand-selected learning resources that we regularly recommend to both beginners and experts.

paper oak
lapis sequoia
#

Hi everyone, my pytorch ai wont work and i've narrowed its issue down to the fact that it doesnt respond to reward, i've attatched the py file of which i use to calculate its next move and how it uses reward for that
https://paste.pythondiscord.com/hekedehepe
and im wondering what i can do to make it respond to reward and calculate a move based on its reward

#

Hello, I just created a bar plot with the following picture, it turns out that there is a text hit the top spine(the red sign), and I want the bar is close to the side spines(the green sign). Does anyone know the solution for these problem? Thank you

IMG_20220509_022502_764.png
#

This is the code

IMG_20220509_022508_683.png
lapis sequoia
arctic wedgeBOT
#

Hey @lapis sequoia!

It looks like you tried to attach file type(s) that we do not allow (.zip). We currently allow the following file types: .gif, .jpg, .jpeg, .mov, .mp4, .mpg, .png, .mp3, .wav, .ogg, .webm, .webp, .flac, .m4a, .csv, .json.

Feel free to ask in #community-meta if you think this is a mistake.

#

Hey @lapis sequoia!

You either uploaded a .txt file or entered a message that was too long. Please use our paste bin instead.

lapis sequoia
#

all set-up

mild dirge
lapis sequoia
lapis sequoia
mild dirge
#

and try not to hardcode the limits, just add like a proportion from the highest bar

lapis sequoia
bronze fiber
#

can anyone suggest any full length pandas tutorial which goes from beginner to advanced topics?

zenith panther
#

hi, i had this code that can scrape data from multiple pages, i tried to run it on google colab but it couldnt cause of the RAM somehow i tried to upgrade it but still couldnt run it so i decided to run the same code in Jupyter notebook and it's still running for 4 hours now . i didnt understand the issue, can someone help me out with it ? thank you in advance.

tacit basin
lapis sequoia
zenith panther
tacit basin
#

So hours doesn't seem right

lapis sequoia
#

does anyone know how to set the linewidth of the spines?

tacit basin
#

Where did you run it when it took minutes?

zenith panther
tacit basin
#

If it's not the same code

#

Difficult to investigate this with so little details

zenith panther
#

so when i try to do the pagination it stuck

lapis sequoia
zenith panther
lapis sequoia
#

infinite loop can be the problem too

lapis sequoia
zenith panther
lapis sequoia
brave osprey
#

hi i want to develop a fintech do you ahve some code

#

can anyone help me with a bayesian hidden markov chain in rule?

#

in chain is the thingh

#

with stan

fiery dust
# 
json.decoder.JSONDecodeError: Expecting property name enclosed in double quotes: line 1 column 2 (char 1)
#

ok apparently if I send a json in this format:

{'symbol': 'AVAXUSDT',  'timeframe': '1',  'side': 'SHORT',  'order_type': 'Market',     'order_price': '52.535',    'tp_min': '52.535',  'tps': '[52.515, 52.51, 52.5, 52.475, 52.45]',  'sl': '52.565', 'rr': '2.8483333333'}

the error appears, but if I send the json like this:

{
   "symbol":"AVAXUSDT",
   "timeframe":"1",
   "side":"SHORT",
   "order_type":"Market",
   "order_price":"52.535",
   "tp_min":"52.535",
   "tps":"[52.515, 52.51, 52.5, 52.475, 52.45]",
   "sl":"52.565",
   "rr":"2.8483333333"
}

the error doesnt appear

echo summit
vestal gale
#

Getting started with ML and AI, would love to see what you guys use for predicting/forecasting sales. Anyone able to help?

misty flint
#

bc idk if you can train a feasible model using speech recognition and word probabilities

#

with NLP

#

like

#

it sounds like the accuracy wouldnt be good but thats just my intuition

safe elk
misty flint
#

bruh

#

really?

#

wouldnt you need really good mic quality

delicate apex
safe elk
#

In paper they say detect 41.8% of keystrokes and 27% of typed words correctly in such a noisy environment---even without user specific training. To investigate the pote

#

Be noisy lmao

delicate apex
#

yeah, keyboards are rather awful security wise, but the vast majority of exploits require physical access, and if a bad guy can touch your computer, the keyboard is not your biggest problem

misty flint
iron basalt
#

There are more fun physical hacks, like using a laser to shine into a printer's LED through the window and uploading a virus.

#

(LEDs are inputs too)

#

It's a non issue though. Just an interesting research project.

misty flint
#

everything is a vulnerability

arctic wedgeBOT
#

Hey @barren wedge!

You either uploaded a .txt file or entered a message that was too long. Please use our paste bin instead.

barren wedge
upper spindle
#

how do I locate my ipynb file, my os is windows 10

lapis sequoia
lapis sequoia
odd meteor
mighty spoke
#

anyone know how to do this

supple leaf
#

Hello, how do I reach the hours in this column?

unknown.png
# 
df['hour'] = [x[11:13] for x in df.index]

It doesnt work.. it says this

odd yoke
#

You must slice the timeperiod column, not the index of the dataframe

supple leaf
#

so it should say df.Timeperiod?

odd yoke
#

Yes

upper spindle
#

does this mean they're in zip files now?

unknown.png
unique flame
#

Anyone here familiar with YOLO? I've used inference in DarkNet and OpenCV on the same image, but the inference on OpenCV has some slightly more detections. Wonder if anyone has experienced something similar.

odd meteor
# supple leaf Hello, how do I reach the hours in this column?

You could easily convert the column to datetime object and extract the hour.

pd.to_datetime(df['time_period']).dt.hour

You could also try using the apply function + lambda on the column to extract the hour as well.

df['time_period'].apply(lambda x: x[10:12])
serene scaffold
#

also wouldn't df['time_period'].str[10:12] be more idiomatic?

odd meteor
serene scaffold
supple leaf
#

Big thanks guys, will try it now

urban lance
#

Hey, thanks for your answer. Sorry I haven't replied till now.
I am having some trouble with the .rank() function. Can't seem to figure it out. also the code you shared is not filtering the first 2 distinct IDs but its only returning 1 unique value

urban lance
#

seems to have a better result but it's still not quite right

tough bolt
#

How do I approach explaining Graph Neural Networks to a group of people who have absolutely no idea about neural networks let alone programming?

wooden sail
#

if they're good at math, you can explain the math

urban lance
fervent flicker
#

if you're* good at math you can explain the math

#

sorry to butt in

urban lance
#

they won't understand a thing either 😛

fervent flicker
#

you just explain all the necessary prereqs

#

if you haven't practiced explaining things enough times it will be hard to explain it all concisely enough

#

and that's why people love cheatsheets

#

there's one on graph theory for nns, i bet

#

and i agree with @urban lance's idea that the human brain is a good analog for graph nns because we have compartmentalized brains that operate like them.

pseudo wren
#

does anyone here have the time to review a brief colab notebook

#

like i'm talking 20 cells

#

half of which are just things you can ignore like setting up the dataset

#

i just need feedback and to see what i'm doing right/wrong

#

notebook in question

candid pollen
#

anyone familiar with lstm? i have a question

tough bolt
tough bolt
spiral iris
#

Any have to read book about DS, AI or smth?

#

To understand the basics,to extend knowledge, to see how to implement effective algorithms, projects for AI on Python?

manic heron
#

recently did some citizen datascience using polars

#

here is the notebook, if anyone would like to give feedback

#

i'm not a pro data scientist, but i'm doing my best

lapis sequoia
#

Is calculus really necessary to understand ml

#

?

spiral iris
#

Calculus is necessary in programming after all

#

Not everything but some parts like integrals are very useful

#

And matrix math

lapis sequoia
#

Oh

spiral iris
#

As far as I know

#

Those are easy parts of calculus

lapis sequoia
#

Okay thanks

#

Linear algebra is the basic for ml right?

lapis sequoia
manic heron
#

absolutely. i think of it as the spiritual successor to pandas

#

it's the future imo

#

i like pandas fine, but i find polars cleaner as an API and faster

#

and the code that i write is just way more elegant and readable

lapis sequoia
#

Oh

serene scaffold
manic heron
#

imo

serene scaffold
spiral iris
manic heron
#

well expressions make a huge difference but:

#

groupby/agg is cleaner

serene scaffold
#

also, does it support sets?

manic heron
#

select/filter are cleaner

lapis sequoia
manic heron
#

i mean these are judgments

#

but what i'd recommend is just rewriting one of your existing notebooks in polars

#

i had a "wow" moment where i was like holy shit, this is just a better way

serene scaffold
#

I actually avoid notebooks as much as possible 👀

manic heron
manic heron
spiral iris
manic heron
#

assuming other people have the same questions as you

serene scaffold
#

thanks for letting me know!

spiral iris
#

Can u explain for us in plain words which instruments did you use?

misty flint
#

thats fair

#

theyve def been abused

#

but they do have their use cases

#

such as rapid experimentation

serene scaffold
spiral iris
#

U say that python > jupyter?

#

Then why there are a lot more jupyter ai projects,then python?

wooden sail
#

jupyter runs python code. it's just that it also allows you to put in tex, markdown, and images all in the same place, too, but you could just as easily make a script in a file (or a few) and run it from the terminal or your favorite IDE. jupyter is one way of doing that, too

spiral iris
#

Oh so jupyter isn't necessary

wispy coyote
spiral iris
#

And I can use a pure python

#

Community version is enough for AI stuff, or I'l have to buy a professional?

wooden sail
#

community version of? python is free

spiral iris
#

Pycharm

wooden sail
#

you don't need pycharm either. the community version would be fine, if you like it. you can pay for the pro version if it has anything you like. technically, all you need is notepad or any other text editor, and to download python

spiral iris
#

There ide better then pycharm as far as I know

spiral iris
wooden sail
#

sure. i don't think windows' notepad supports syntax highlight though, which is imo the core thing you'd want. you could use, for instance, notepad++ or something like that

#

or spyder, if you like it. many people like vscode

spiral iris
#

I'm on linuxpy_guido

#

Pycharm isn't a bad choice imo

wooden sail
#

on linux, your default text editor supports syntax highlighting

#

or you can use vim, emacs, or (the one i use) micro

spiral iris
#

You also are on linux?😎

#

I heard that vim takes ages to learn how to use it properly and to get accustomed to it's features

serene scaffold
wispy coyote
manic heron
manic heron
serene scaffold
manic heron
serene scaffold
#

Linear execution order, being able to import stuff, modularity

#

Notebooks are fine for exploratory analysis. But they're bad for production.

#

I mostly use an IPython repl for quick experimentation though.

manic heron
#

you can run a notebook in order too

#

you can import stuff

serene scaffold
#

I mean being able to import stuff that you create

#

if you define a function in a notebook, how do you import that into something else?

manic heron
#

idk. it seems like you have a strong opinion about this, and i won't try to change your mind, but it is possible to import functions defined in notebooks

#

we used notebooks in a pretty serious cross-lab collaboration, it didn't seem to be an issue tbh

#

but yea, everyone has their tastes

serene scaffold
#

how do you import functions defined in a notebook?

manic heron
#

if that was a serious, non-rhetorical question

agile cobalt
#

I for one use Atom+Hydrogen, which allows for me to run python code somewhat like IPython but in normal .py files - that way I can test stuff while still keeping it well formatted

manic heron
#

yea you can do that with vscode as well

#

it's dope

agile cobalt
#

I admit that importing stuff can be a pain though, so breakpoints + to/from clipboard is useful

lapis sequoia
#

I use atom. And ran it earlier in the regular python console. Now I do it with vscode.

agile cobalt
lapis sequoia
#

Damn. Looks like it is a transformation of atom into vscode

#

What's better in life 😁

echo vigil
arctic wedgeBOT
#

@echo vigil :white_check_mark: Your eval job has completed with return code 0.

001 | Before:
002 |     user_id  timestamp auxillary
003 | 0        1          1         a
004 | 1        1          2         b
005 | 2        1          3         c
006 | 3        1          4         d
007 | 4        2          7         e
008 | 5        2         10         f
009 | 6        3          0         g
010 | After:
011 |     user_id  timestamp auxillary
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/xajiseride.txt?noredirect

echo vigil
#

Correct me if I'm wrong but that seems to be working fine -- what should be different about the output on this test case?

urban prism
#

Is there way I can make changes on installed libraries on Kaggle?

misty flint
desert oar
desert oar
#

https://dvc.org

Data Version Control · DVC
Data Version Control · DVC

Open-source version control system for Data Science and Machine Learning projects. Git-like experience to organize your data, models, and experiments.

misty flint
desert oar
#

I can't speak to how it would scale across a large team, but I've used it in a small team and it worked pretty well

#

It also made sharing data and trained models pretty easy, less time wasted re-computing things

fervent flicker
#

it's a project related to dvc. a trained model database, if i understand correctly

nocturne wigeon
#

can someone help with an exe of a bot i built please?

serene scaffold
nocturne wigeon
#

channels quotes "ai"

#

and it is ai

serene scaffold
#

okay, so it's not a discord bot? what is the problem?

nocturne wigeon
#

no not a discord bot

#

a chess bot

#

trying to figuring out it's language, i tried like 3 decompilers

#

maybe its py

#

not sure tho

#

its an program but i forgot what language it is written in

serene scaffold
nocturne wigeon
#

mhmh well not what im looking exactly for but i'll give a shot ok

eager cloak
#

How would I fix this?

unknown.png
#

The code: ```py
from bitcoin import *
import bitcoinlib

def new():

public_key=privtopub(private_key)

private_key = random_key()
public_key=private_key.to_public()
address=public_key.to_address('P2PKH')
P2WPKH=public_key.to_address('P2WPKH')
print(f'Priv key>{private_key}\n Pub key> {public_key}\n Address>{address}\n P2WPKH> {P2WPKH}')

new()

lapis sequoia
#

Not sure if this would be the right channel but

# 
from ib_insync import *
from numpy import *
import yfinance as yf
import time
import pandas as pd 

ib = IB()

#vars
contract = Stock('MSFT', 'SMART', 'USD')
nextOperation = True #True = buy, False = sell
#Buys the asset if its price decreased by more than the threshold.
dipThreshold = -2.25
#Buys the asset if its price increased by more than the threshold.
upTrendThreshold = 1.5
#Sells the asset if its price has increased above the threshold since bought.
profitThreshold = 1.25
#Sells if its price has decrease by more than this threshold to stop loses.
stopLossThreshold = -2
def returnCurrentPrice():
    stock = yf.Ticker('MSFT')
    return stock.info["regularMarketPrice"]
def returnHistoricalPrice():
    stock = yf.Ticker('AMD')
    return stock.history(period='7d', interval='1d')
lastOpPrice = 90
buyOrder = LimitOrder('BUY', 1, returnCurrentPrice())
sellOrder = LimitOrder('SELL', 1, returnCurrentPrice())
def returnBalance():
    accountBalanceString = ib.accountSummary()
    for a in accountBalanceString:
        if a.tag=="CashBalance":
            return double(a.value)

#funcs
def attemptToMakeTrade():
    precentageDiff = (returnCurrentPrice() - lastOpPrice)/lastOpPrice*100
    if nextOperation == True:
        tryToBuy(precentageDiff)
    else:
        tryToSell(precentageDiff)
# 
def placeBuyOrder():
    print("Placing buy order")
    ib.placeOrder(contract, buyOrder)
    lastOpPrice = returnCurrentPrice()

def placeSellOrder():
    print("Placing sell order")
    ib.placeOrder(contract, sellOrder)
    lastOpPrice = returnCurrentPrice()


def tryToBuy(precentageDiff):
    if precentageDiff >= upTrendThreshold or precentageDiff <= dipThreshold:
        placeBuyOrder()
        nextOperation = False
        print("Next operation is", nextOperation)

def tryToSell(precentageDiff):
    if precentageDiff >= profitThreshold or precentageDiff <= stopLossThreshold:
        placeSellOrder()
        nextOperation = True
        print("Next operation is", nextOperation)

def startBot():
    print("Starting...")
    ib.connect('127.0.0.1', 7497, clientId=1)
    lastOpPrice = returnCurrentPrice()
    nextOperation = True
    while(1 > 0):
        attemptToMakeTrade()
        time.sleep(30)

startBot()
#

Its expected to change the nextOperation bool but instead just repeats the tryToBuy and placeBuyOrder

#

and it just repeats that non stop

#

sorry if its a bit of a text wall

arctic wedgeBOT
#

:incoming_envelope: :ok_hand: applied mute to @lapis sequoia until <t:1652142105:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

lapis sequoia
#

lmao

arctic wedgeBOT
#

pytorch_grad_cam/base_cam.py line 62

def forward(self,```
`pytorch_grad_cam/base_cam.py` line 74
```py
outputs = self.activations_and_grads(input_tensor)```
`pytorch_grad_cam/activations_and_gradients.py` line 39
```py
def __call__(self, x):```
urban prism
#

This bot is amazing.

arctic wedgeBOT
#

No, you are lemon_starstruck

strange stag
#

looking for some help with ray and custom environments
code is bare-bones and very very simple, and the environment is as simplest ive ever seen
entire traceback is also included
https://bpa.st/F5SA

strange stag
#

hmm, seems i needed to follow the directions on the ray website and add the env_config param to the env __init__

#

however, getting some numpy errors now 😕

novel python
#

Guys, how do you change the figsize for jointplots and pairplots with Seaborn? I tried going for the usual plt.figure(figsize=(12,6)) but it clearly doesn't work, couldn't find anywhere how.

wind jay
#

im trying to figure out how i would get started with artificial intelligence, say for example with text. how might i go about doing that?

#

like is there any good tutorials, websites anything reallly

#

would you mind expanding on that?

tacit basin
tacit basin
# wind jay would you mind expanding on that?

So this is one of the ways and it's a practical way, you learn stats, and all else on a learn as you need basis. So you can start playing the real game, NLP in your case as soon as possible. With the current libraries like hugging face you can get a lot done with minimal code or stats. Then you learn as you go.

tacit basin
#

is python 3.10 supported by major DS libs? like pandas, scki-kit, matplotlib, pytorch, etc. ?

left hazel
#

Sadly not yet by pytorch (soon!), but most others do.

eager cloak
#

Heya, so this code is giving me the error belowpy from bitcoin import * import bitcoinlib def new(): private_key = random_key() public_key=private_key.to_public() address=public_key.to_address('P2PKH') P2WPKH=public_key.to_address('P2WPKH') print(f'Priv key>{private_key}\n Pub key> {public_key}\n Address>{address}\n P2WPKH> {P2WPKH}') new()
Error: Desktop>python main2.py Traceback (most recent call last): File "main2.py", line 12, in <module> new() File "main2.py", line 8, in new public_key=private_key.to_public() AttributeError: 'str' object has no attribute 'to_public'

#

How can I fix that?

raven torrent
#

what is the best ML library that is user friendly that is not tensorflow or keras

#

@eager cloak what is the best ML library that is user friendly that is not tensorflow or keras

low bronze
#

hi

#

hi, I get a warning when i do this but it works, I dont know the right way to do this

for col in  df.columns[1:]:
    df[col] = pd.to_numeric(df[col], errors='coerce')

I get this warning. Could not find a way to solve this on stack over flow

/local_disk0/tmp/1652161177867-0/PythonShell.py:147: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
self.io.write(decoded)

#

please @ me

celest vine
#

I have a supervised classification problem (yes,no), but the data is highly imbalanced. Yes - 5% and No - 95%.
Is it possible to make a good prediction model using this data?

tacit basin
urban lance
#

this is the result I'm looking for:

a | 1.0
a | 1.0
a | 1.0
b | 2.0
c | 3.0
c | 3.0
d | 4.0

a count of distinct values

eager cloak
tacit basin
tacit basin
eager cloak
celest vine
desert pine
#

Hi, I have a question. Is that possible to create product recommender systems based on their phone number instead of their username on website since
my sales dataset mostly came from whatsapp

urban lance
sudden egret
#

Have anyone worked on Time Series Classification? I am working on a problem now where I have to forecast the change in insurance carrier.

The inputs I have now so far is the Insurance Carrier Name, Companies and Premium amount of each company.

So the goal is to build a model that takes these inputs and forecast whether the company would change their carrier in future or not.

Still now I tried windowing techniques but I feel like Im missing out somewhere. There are almost 99 different companies.

#

Any help or tidbits to get started on this would be highly appreciated.

#

This is how the sample data looks.

Screenshot_2022-05-10_at_2.50.28_PM.png
tawny vine
#

how to create an ai voice assistant

urban lance
low bronze
robust charm
#

Hi, has anyone here used stablebasline3?

tawny vine
tawny vine
robust charm
# tawny vine yes

Hi im training a chess AI using self play. I have a couple questions about rewards and training if you have the time

lapis sequoia
#

hey guys, How to combine multiple hyperspectral images into one? ```py

Example: Iterate over data set

for sample in dataset:
datacube, labelmap = sample
print(datacube.shape, labelmap.shape)
output -
(240,520,16)(240,520)
(240,520,16)(240,520)
(240,520,16)(240,520)
(240,520,16)(240,520)

charred spear
queen torrent
sudden egret
queen torrent
inland zephyr
#

does anyone have good references for document verification methods? i wonder if non-CNN approach is available somewhere

serene scaffold
inland zephyr
#

just if a word or sentence is in the document

#

like let said the sender name should be inside the document

#

i know there is possibility that the name in the document will be abbreviated, so i will try to check the similarity of the phrase in the document with the sender name

odd meteor
# inland zephyr like let said the sender name should be inside the document

If you already knew the specific words/sentence you wanna search for in the document, you could do something like this


list_of_words = ['Money Laundering', 'Police', 'Weed', 'Cocaine', 'Drugs', 'Gun']

df.loc[df["document_message"].str.contains('|'.join(list_of_words), na=False)] 

'''
If you wanna flag 🚩 document where such words appeared you could do this
'''

df['flag'] = np.where( (df['document'].str.contains('|'.join(list_of_words)) == True) 1, 0)
misty flint
#

nice word list

serene scaffold
odd meteor
lapis sequoia
#

!docs

arctic wedgeBOT
lapis sequoia
#

!resources

arctic wedgeBOT
#
Resources

The Resources page on our website contains a list of hand-selected learning resources that we regularly recommend to both beginners and experts.

lapis sequoia
#

can anyone give me ML docs

serene scaffold
urban prism
#

Does someone here have experience with OpenMMLabs?

inland zephyr
mental bane
#

This might be a dumb question but can someone please explain me the use of 'passthrough' in this code and in general?

unknown.png
tacit basin
urban prism
#

Do you know how img_metas should be formed? I'm using another library and it sends the image to mmdet without that, which causes an error. I changed some lines in the libraries to get past it but now I have to manually declare it in the library's code

arctic wedgeBOT
#

mmdet/models/detectors/base.py line 112

def forward_test(self, imgs, img_metas, **kwargs):```
gray orchid
lapis sequoia
urban prism
#

What ML library? There are many of them @lapis sequoia

serene scaffold
lapis sequoia
#

oh idk about ml lib let me check it out

serene scaffold
#

It sounds like there's a misunderstanding here. What are you trying to do? Just learn about machine learning?

lapis sequoia
#

im confused

#

um these?

unknown.png
#

can u suggest which one should i use? irdk

neon imp
#

Not sure this is the channel, but anyone done any quantum computing work?

serene scaffold
serene scaffold
lapis sequoia
#

okay

misty flint
neon imp
#

Yeah I was just wondering. Quantum has a lot of overlap with Data Science stuff in my head, was wondering if some people here did it.

#

Just because Quantum is 100% about state following a probability distribution.

median moat
#

Everywhere I go I see his(rex) face.

misty flint
median moat
#

No don't leave

plucky harness
#

I need some help. I have created a handwritten data of 1000 png images and now I'm stuck with the part of pre-processing and feature engineering. I mean I am facing problems with how to import, load, and process the data for model training with scikit-learn.

mighty bloom
#

someone help me in finding the syntax error in this?

unknown.png
tacit basin
tacit basin
mighty bloom
mighty bloom
tacit basin
mighty bloom
tacit basin
#

from plotly website i see this should be the syntax?:

@app.callback(
    Output(component_id='body-div', component_property='children'),
    Input(component_id='show-secret', component_property='n_clicks')
)
mighty bloom
mighty bloom
gray swallow
#

Guys can you help me

unknown.png
main geode
#

Hey I am looking for some amazing people to colab on a project, anyone's interested.

gray swallow
#

I'm using SVM classifier

unknown.png
#

I used two columns as outputs (Y) for my model

main geode
#

I am a beginner in ML and wanted to begin with some good project if we could make up a team or someone could add to his project would love to help

gray swallow
main geode
#

@gray swallow y ( the label) should be a 1dimension array

eager cloak
#

How can I turn a P2PKH BTC address into a P2SH address?

tacit basin
gray swallow
tacit basin
gray swallow
main geode
#

what dataset are you using

gray swallow
#

Here is the dataset. I combined risk_score and default values as outputs . So I zipped them into a list of tuples and did the same for the X values

📎 financial_data.csv
#

import pandas as pd
import matplotlib.pyplot as plt

import seaborn as sn

import numpy as np

from sympy import together

import matplotlib.pyplot as pyplot
import pickle
from matplotlib import style
from sklearn.model_selection import train_test_split
from sklearn import svm
from sklearn import metrics

import data

dataset = pd.read_csv('financial_data.csv')

now we start the Exploratory Data Analysis process

print(dataset.head())

response = dataset["default", "risk_score"]

response values

default_values = dataset["default"]
risk_score_values = dataset["risk_score"]
place_holders = np.zeros( len(risk_score_values))

print(place_holders)

exit()

Y_values is a combination of risk score and default values

Y_values = list(zip(risk_score_values, default_values, place_holders, place_holders, place_holders, place_holders, place_holders, place_holders))

Y_values = list(zip(risk_score_values, default_values))

X values defined here

age = dataset["age"]
pay_schedule = dataset["pay_schedule"]
home_owner = dataset["home_owner"]
income = dataset["income"]
months_employed = dataset["months_employed"]
years_employed = dataset["years_employed"]
has_debt = dataset["has_debt"]
amount_requested = dataset["amount_requested"]

zippping X values together

X_values = list(zip(age, pay_schedule, home_owner, income, months_employed, years_employed, has_debt,amount_requested ))

drop columns

dataset = dataset.drop(columns=["default","entry_id", "risk_score"])

X_train,X_test, y_train, y_test = train_test_split(dataset, Y_values, test_size=0.2, random_state=0)

clf = svm.SVC()

clf.fit(X_train, y_train)

y_predict = clf.predict(X_test)

acc = metrics.accuracy_score(y_test, y_predict)

print(acc)

#

That was the code

main geode
#

No bro that's now how svm works, It requires a label of 1d array not combination of all the col

#

choose one feature as your label and then make predictions

gray swallow
main geode
#

Here i can see Risk_score is the label here, therfore y = dataset["risk_score"]

#

Your code would work more nice if you use jupyter notebook or google colab

#

you can do the process step by step though jupyter notebook

gray swallow
#

Ok thanks bro

#

I'm creating a system to predict whether any applicant would default or not.

#

Will it work if I use default as my label here

main geode
#

yes definitely

barren wedge
#

How to solve this?

RuntimeError: CUDA out of memory. Tried to allocate 5.56 GiB (GPU 0; 11.17 GiB total capacity; 6.59 GiB already allocated; 3.88 GiB free; 6.73 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

eager cloak
#

All the text is styled in yellow, is there a way for me to seperate it?
Like make the unblurred text yellow and the other text green

This is essentially the code```py
from bitcoinaddress import Wallet
info=input(f"{bcolors.RESET}Private key>{bcolors.YELLOW} ")
wallet = Wallet(info)
print(wallet)

unknown.png
plush glacier
#

does anyone here know how i could add a tensor of shape (3,4) to a specific part of a larger tensor for example a tensor with shape (40,10) and add it to the point of (20:3,3:4) of the large tensor
with tensorflow
it is a small part of a school assingment the entire assingment is making a tetris ai and this is just for the tetris game part

#

pls ping me if you reply to me

wooden sail
tacit basin
hasty mountain
#

Hey guys, can someone give me some help with Gym-Retro? I didn't edit not even a single thing in the files, yet I still get this error.

unknown.png
#

This is (literally) the entire code(excluding imports):

env = retro.make(game="StreetFighterIISpecialChampionEdition-Genesis")
obs = env.reset()

env.render(close=True) # habilita o fechamento da janela
env.close()

checkpoint = CheckpointCallback(save_freq=100000, save_path="D:/Python/Projects/Hakisa")

model = PPO2(policy="CnnPolicy", env=env, gamma=0.99, n_steps=64, learning_rate=3e-9, vf_coef=0.5, verbose=1)
start = time.time()
model.learn(total_timesteps=1000000, log_interval=100, reset_num_timesteps=True, callback=checkpoint)
end = time.time()

print("Duration: ", (end-start)/3600)
#

PS: Ping me if you can help.

lapis sequoia
hasty mountain
#

Unless that garbage collector is deleting the current emulator I'm running, but this shouldn't be happening, right?

barren wedge
tropic pecan
#

Hi Yall, I'm working on a project where we are trying to optimize some cooling patterns for a specific manufacturing process and the simulation result will be similar to this image and I also have the data with coordinates and temperatures at those certain points during that process. So I was wondering what sort of ML algorithms I would need to optimize the cooling patterns

unknown.png
plush glacier
wooden sail
#

all right. well the main issue is that tensorflow likes keeping things const

plush glacier
#

ik so is there a way to add it to it

wooden sail
#

one work around is to circumvent it entirely: cast your tensors to numpy nd arrays, do the math there, and then make a new tensorflow tensor and assign it to the original one

#

alternatively, you can make tensorflow modify the entries of the tensor, but i don't recall how to do that in any easy way

plush glacier
#

wouldn't it be possible to pad the tensor that is added

#

to match the size

wooden sail
#

sure, that's also an option, but that's very inefficient

plush glacier
#

well it is only for very small tensors so memory isn't a issue

#

it would probably be like 10x faster than the numpy solution

wooden sail
#

probably, yeah

plush glacier
#

because the data doesn't have to be moved from gpu to cpu and then back to the gpu

#

not that that would be a solution that i would want because i dont know how it will pad the image

wooden sail
#

how about this

#

this works, for example:

import tensorflow as tf
import numpy as np

x = tf.constant( [[1,2,3],
                [4,5,6],
                [7,8,9]])

y = tf.constant( [[1,1],
                [3,3]] )

indices = np.zeros((2,2,2), dtype=int) #2 x num rows in y x num cols in y
indices[0,:,:] = np.arange(0,2).reshape(-1,1) #row indices
indices[1,:,:] = np.arange(1,3).reshape(1,-1) #col indices
indices = indices.reshape(2,-1).T #reshaped into size #inds x 2
print(indices)

print(tf.tensor_scatter_nd_add(x, indices, tf.reshape(y, [4])))
#

pretty convoluted though, and does not add in-place. it makes a new tensor

proud pond
#

Hellow everyone
I tried a lot of times to learn reinforcement learning, but I always find myself lost or I don't know what's the next step or what to learn.
I would like to put a plan for myself to learn so I don't find myself one day lost and without a task.
Can you guys help me with that, would appreciate it.
Like with resources, a link, a plan or anything.
Thank you pixels_snek_2

tacit basin
plush glacier
#

it seems to support a bit more things

#

and no numpy needed

#

although i dont know if i could make a solution that jit compiles well

wooden sail
#

no numpy was needed in what i did either. this thing also makes the tensor sparse, so the 0 entries don't wase more memory

#

what i did was just a makeshift cartesian product. there are other ways of doing it without numpy

#

and it was only to make the indices, not the actual tf tensors

plush glacier
#

oh to place the thing at the right place i think it will be just some simple math

wooden sail
#

probably the same math i put there 😛

#

numpy and tensorflow (probably) should have a meshgrid function that does the same thing

plush glacier
wooden sail
#

there you go

#

but this will make the index array in gpu, which i thought was kinda pointless

plush glacier
#

i dont think that a index array would be possible because each game will be slightly different

wooden sail
#

you need the index array both for padding and for the function i suggested

#

otherwise you don't know where in the larger tensor to add the smaller one

plush glacier
#

it is always at the same place at the start

wooden sail
#

sure, and then it will presumably change as time goes on. you have to track it some how

#

either in gpu or memory

plush glacier
#

well my solution for that was to make it 0.5 instead of 1

wooden sail
#

make what 0.5?

plush glacier
#

the value of the one that can move

wooden sail
#

idk what you mean by that. if it works for you, cool 😛

#

but in any case, you always need to know WHERE to add the smaller tensor, otherwise the operation is not well defined. neither in math nor in code

plush glacier
barren wedge
safe moss
#

is this the channel to talk about pandas in?

rose agate
#

the animal? no. the package? yes

safe moss
#

excellent

#

i have a query,

i had three dataframes that i believe i correctly processed so that their columns were identical and any that were not present in all three were removed

#

as you can see they were all reduced to 116

unknown.png
#

but i used the concatenate command on them, expecting them all to merge seamlessly and remain at 116 columns but i have experienced this unexpected result and i am not sure why

#

( i added one extra column)

unknown.png
#

the display comes out with 233 columns and i am not sure why

unknown.png
rose agate
safe moss
rose agate
#

what is the contents of q_set?

safe moss
#

so basically all three data frames had a number of columns, i then reduced them down so that the only columns left in each were columns that were present in all three dfs

# 

for df in df_array:    
    for question in df.loc[0]:
        map[question] = map.get(question, 0) + 1

q_set = set()

​

for key, val in map.items():

    if val == 3:

        q_set.add(key)

print(len(q_set))

116
rose agate
#

Are there null values in the columns of the final frame?

rose agate
#

You might need to specify axis=1

safe moss
#

i assume you mean in the newly created df?

rose agate
#

Other than that I'm not sure

#

Yeah

safe moss
#

one moment

rose agate
#

Also I'd try if set(df1.columns) == set(df2.columns)

safe moss
#

what will that do, sorry my brain is a bit fried atm from something else

#

creating two sets and comparing them?

rose agate
#

Yes, just testing that the columns in each of the data frame are the same

#

Also I'd try adding axis=1 in your statement and see if that fixes it

safe moss
rose agate
#

Either your problem is that the columns aren't the same names or something else I don't know

safe moss
#

ill try the set

safe moss
rose agate
#

Looks like it, try getting q_set by doing the intersections of those sets

#

Not sure if I'm remembering right, but I think it may only allow you to do 1 intersection at a time

versed gulch
#

Hi does anyone know how to overlay 2 images over one another in python i.e. the mask onto the original image using the numpy arrays?

unknown.png
tacit basin
wooden sail
safe moss
rose agate
versed gulch
safe moss
#

do you mean doing concat as per before or changing the formulae

wooden sail
rose agate
#

I think axis=1 was wrong, try without

versed gulch
rose agate
#

Make sure you update the frames with the function you wrote before to drop everything not in the set

lapis sequoia
#

So i have to switch rows and columns without using the transpose function. This code is somehow not working, and is giving errors because I use matrix[y] and matrix[x][y]. I dont get why, does someone know how to fix this code?

2022-05-11.png
wooden sail
#

what error do you get?

#

also, what type is "matrix"? list of lists? np array?

lapis sequoia
#

matrix is a pandas dataframe

wooden sail
#

ah. sadly i've never used pandas :x hopefully someone else can help you out

lapis sequoia
#

thank you anyways!

serene scaffold
lapis sequoia
#

for this function i'm not allowed to use df.T. I am trying to do it with a for loop but it's giving errors

serene scaffold
lapis sequoia
#

yes it's an assignment

serene scaffold
#

alright. Can you show the code as text in a markdown block?

#

!code

arctic wedgeBOT
#

Here's how to format Python code on Discord:

```py
print('Hello world!')
```

These are backticks, not quotes. Check this out if you can't find the backtick key.

serene scaffold
#

^ this is the only format I will accept. I will also not accept screenshots of error messages.

lapis sequoia
# 
def transpose(matrix):
    transposed = pd.DataFrame()
    
    for y in range(len(matrix)):
        for x in range(len(matrix[y])):
            transposed[y][x] = matrix[x][y]
    

    return transposed
wooden sail
#

my impression is that, instead of a double loop, you'd wanna iterate through the rows or columns only. pick one, not the other. read the lists, and put them into another list of lists. then make a new dataframe from that. i don't know the exact functions to do that though

serene scaffold
#

There are a few problems here:

  • You would need to initialize the DataFrame with empty cells for the desired shape
  • You need to also transpose the columns and indices of the original DataFrame and put those in the transposed DataFrame
  • You index Dataframes with the iloc accessor, which means "index location". See the next point:
  • DataFrames are one data structure. They are not a list of lists. Expressions like transposed[y][x] should be transposed.iloc[x, y] because you are indexing the one data structure, not two.
#

Keep in mind that you would never do this in a real situation, and this assignment is almost completely pointless.

#

any "real" code that involves allocating empty space in a DataFrame and then putting stuff into it later is wrong.

#

oh, you also need to be using .iloc

lapis sequoia
#

okay, i guess that should help me for now, thank you! It's indeed a really annoying assignment since df.T would be so much easier and faster ;/

serene scaffold
brazen totem
#

Sorry to barge in but I was told I should ask here instead of the general question channel

Let's say I'm trying to predict a category (decided using arbitrary statistical cutoffs) using these same cutoffs with other additional statistics. How do I determine how well my regression predicts a category without arbitrarily assigning them beforehand?

Or is it just impossible?

serene scaffold
brazen totem
#

I'm working on a project on categorizing poker player types. I use 3 statistics VPIP, PRF, and Limping to decide which player is which category. If these 3 values are within certain ranges they are marked as that player type. Then I run an aggregate on each category onto a larger set of statistics. Then I use both these aggregate statistics and the 3 statistics I used to sort category to predict player category

#

I'm trying to see if this makes sense or I'm doing something really wrong

serene scaffold
#

so you are trying to decide what category a poker player belongs to. in data science we call this "classification". so you care trying to predict the classes of poker players.

brazen totem
#

Yes

serene scaffold
#

well, I guess it might not be classification per se if the class is determined by properties of the player that are always known

#

anyway

wooden sail
#

it does make sense. you could take the statistics as a vector of numbers input into your classifier, and the output is the class the player belongs to (encoded in some way. probabilities, most likely)

serene scaffold
#

nothing about it immediately jumps out at me as terrible

brazen totem
#

The idea is to first create broad categorizations of poker players to create strong counter strategies

and then try to predict which category a random player is given their stats (to choose the proper strategy)

wooden sail
#

the only thing is that depending on how the aggregates are computed, some might be redundant. not a huge problem if it's not that many

brazen totem
#

I'm trying to do this in python but I'm not sure how exactly I would structure this

wooden sail
#

you're essentially trying to pair up vectors of these statistics with labels. how you do that is a bit more freehand, though. for example, the labels could be encoded as a single number or as a vector. i also wouldn't know what the best architecture for this is. a support vector machine can split up the parameter space with hyperplanes, but you probably don't know a priori that this is the best way of doing it

brazen totem
#

it feels like it would just predict with 100% accuracy if I include the 3 main stats I used to categorize in the first place

wooden sail
#

if you generate the numbers yourself perfectly and without any noise, and this model perfectly describes what you are analyzing, yes, that would be the case (and also if the phenomenon can be measured without noise). this is why results that are entirely synthetic should be taken with a grain of salt if you don't put some work into noise statistics

brazen totem
#

so I need to add some randomization into how I categorize?

wooden sail
#

you'd probably wanna contaminate the measured numbers on purpose, or as you say, label some data incorrectly, or something like that. if you're making everything up and there is no noise or anything, and you know the model perfectly, there's hardly any need to use ML in the first place

#

you could use classical optimization methods on your known model to learn the parameters and that's all

brazen totem
#

since it takes an obnoxious amount of hands to be certain about stats

wooden sail
#

that's not really noise, just missing info. the thing is, if you are making up a definition yourself in which you assign a label by hand to some set of vectors using something simple like inequalities, the problem is very easy and you don't need ML 😛

brazen totem
#

so id just use a regression?

wooden sail
#

if you only have 3 parameters, you wouldn't even need that

brazen totem
#

so you think I should go the opposite direction and see how the aggregate stats NOT included in the 3 parameters predicts class?

wooden sail
#

but for the sake of argument, let's say these 3 numbers are indeed noisy with unknown noise stats. and that maybe the labelling process is a bit more complicated. then it makes sense to use ML.

#

well, why would you not include the 3 params?

#

the question i'm asking you is, is the problem so easy that you don't really need ML? there's no point in hiding the parameters "just because"

brazen totem
#

wouldn't they perfectly predict class? since those are the only inputs

wooden sail
#

if there is no noise, yes

brazen totem
#

there is lots of noise

wooden sail
#

so my question goes kind of like this

#

let's say we have numbers a and b in a vector, [a,b]. if a > c_0 and b > c_1, then we assign class A

#

so on and so forth. this gives us 4 classes.

#

next we receive the numbers [a,b] from some scenario we've never seen before. do we know anything about a and b a priori? can we trust the numbers, or do they have mistakes?

brazen totem
#

these numbers have mistakes

wooden sail
#

and going a step further, do we know c_0 and c_1, or not?

brazen totem
#

we define c_0 and c_1

wooden sail
#

all right. then it does make sense. it's a probability estimation problem. how certain one is that the underlying a > c_0 given only a noisy a, and having learned noise statistics from previous examples

brazen totem
#

alright so how would I go about coding that?

#

should I use sample size to determine how noisy a/b are expected to be?

#

and then take into account that noise when assigning class?

#

so instead of saying this player is 100% class Y or class Z

#

we say there's a 40% chance of them being class Y and a 60% chance of them being class Z?

wooden sail
#

i don't think you have an appropriate model to say that though, unless you have good reason to make a probability distribution dependent on those 3 numbers

#

if most of the data is labelled correctly, it should be ok as is

brazen totem
#

poker standard deviation is generally agreed upon? although depending on playstyle it ranges

wooden sail
#

i can't comment on that, idk

brazen totem
#

but if it isn't needed I won't bother with it lol

brazen totem
#

vs correctly

#

yes there;s noise in the input stats but if the input stats are all we use to classsify wouldn't it necessarily be 100% accurate?

wooden sail
#

the only way is to have some ground truth you labeled by hand, then

#

since the labelled data using only thresholds can be incorrect

brazen totem
#

this feels like a chicken and egg problem

wooden sail
#

you can't very it because you don't know the ground truth

#

this is more a clustering problem than a classification one, then

#

there is nothing to very against

brazen totem
#

yeah actual player type is impossible to be certain about

#

we can only have a level of certainty over the stats we have

wooden sail
#

then maybe approach is as clustering instead

brazen totem
#

clustering would be letting the data catogorize itself?

wooden sail
#

pretty much

#

though in fairness, even if the data has a few mistakes in the labelling, it should still work

brazen totem
#

what if it just says nothing though? that there's too much noise to be certain that any player type exists

wooden sail
#

that's a result in itself, then, isn't it?

brazen totem
#

yeah but that's pretty obviously incorrect

wooden sail
#

what i mean is that it would mean the data has no inner structure, and the categories were arbitrary from the get-go

#

if that's the case, there is no difference from simply doing a set of inequalities yourself, no network needed

brazen totem
brazen totem
wooden sail
#

don't worry about it. go ahead and try it out

#

label the data somehow and see if the network learns anything interesting

brazen totem
#

so first id
label player type based on those 3 stats
let machine algorithm try to create its own player types

#

see the correlation between the 2?

#

alright that's something I probably know how to do

#

in that case which clustering method should I use

#

I plan on having 5 player types

wooden sail
#

that's one thing. the other is to also treat it as a classifier problem and see what the network does. you'd expect it modifies the threshold values in some sense. idk how well it will work out with so few dimensions though.

brazen totem
#

Ill just do both tbh

wooden sail
#

that would be the usual thing of taking an image and saying it's a dog, take another image and say it's a cat

#

and feed them to the network

#

i would say to try both approaches, yeah

#

see if you find anything interesting

brazen totem
#

so id include the aggregate statistics into the machine learning model right

#

alongside the main 3 classifiers?

wooden sail
#

you can if you want, why not. you can also test if including them makes any difference, too

brazen totem
#

yeah tbh the one thing ive learned in my economectrics class is that data science is basically trying a bunch of random shit and seeing if it tells you anything

#

😂

wooden sail
#

especially with this type of black box ML you're dealing with here, where the model is based on... not much

#

it gets pretty artistic

#

there are other problems where the architecture and target functions have very good foundations, but i don't think this is one of them 😛 play around and see if anything interesting comes out

brazen totem
#

yeah I'm just trying to see if I can get something useful using hud stats

#

and potentially classify players on the fly

#

using limited/noisy information

misty flint
#

however you usually need an SME for such methods

brazen totem
#

what's an SME?

misty flint
#

subject matter expert

#

for the domain you are working in

#

this is crucial for making sure you are looking at the right features and your assumptions are correct

brazen totem
#

ahhh Id say I know SME

#

I know a player that has gone really deep into this shit

wooden sail
#

right, that's why i was asking you so much annoying stuff about the stats

#

if you could incorporate that, it'd be better

brazen totem
#

how would I incorporate that though?

wooden sail
#

you could turn it into a parametric estimation problem where you update your priors on the distributions

brazen totem
#

use a SME to manually classify each player?

wooden sail
#

basically have ML-powered bayesian estimation

#

if you could assign a parametric family to how each of the numbers you feed the network are distributed, that'd be a good start

brazen totem
#

ah but it depends on the sample size

#

say you have 500 hands vs 50,000 hands

wooden sail
#

you mentioned variance earlier, but my guess is that that is precisely... a guess. that the numbers were assumed to be gaussian distributed

brazen totem
#

the variance will be immensely different

wooden sail
#

the sample variance, yes

#

the true variance of the underlying distribution, no

brazen totem
#

ah I use poker variance calculator

wooden sail
#

the idea is to find the parametric model that best explains the observed data

brazen totem
#

ahhh

#

how would I do that

wooden sail
#

you can read into maximum likelihood and bayesian estimation layer

#

for now i'd suggest to try what we discussed now, and see if it does what you wanted

brazen totem
#

these stats are generally regarded to be normally distributed

#

but I'm not sure why

wooden sail
#

wave your hands wildly while yelling central limit theorem

#

then the name of the game is hyperparameters

#

each of your 3 statistics is contaminated by noise, and so it has its own mean and variance for what you say is usually regarded a normal distribution

brazen totem
#

ah yep

wooden sail
#

the idea is to have the network estimate the mean somehow

#

and then to realize that both the mean and variance have their own... mean and variance

#

if the variance is low, you can trust the result (these are the so-called "error bars" and are related to p values, as some people call them)

#

then you can attach a certainty to your classification

brazen totem
#

ahhhhh

#

that sounds exactly what I wanted to do

wooden sail
#

you can do that with usual classification btw, but that does not consider the statistics of the measured parameters explicitly

#

i would still recommend you begin with what we discussed earlier because this other stuff can get hairy. if you have time constraints, gotta use that time wisely

brazen totem
#

yeah I have literally 24 hours

wooden sail
#

yeah, no

#

do the other stuff

#

do it like a classical image classification problem

brazen totem
#

but this 3rd thing seems super interesting

#

how would I go about doing it? ill do it last if I have time

wooden sail
#

you would need to make a custom cost function based on the maximum likelihood estimation. since this is gaussian, funnily enough, it's the same as doing least squares. however, the knowledge that each parameter is gaussian distributed instead of normal distributed means there is an extra term multiplied to it (added, after taking the log-likelihood) that depends on the statistics of each parameter

brazen totem
#

yeah I have zero idea how to do that lol

#

I've learned about maximum likelihood estimation but I thought it was for binary prediction

wooden sail
#

i would more call it a "philosophy" or solution approach. make a cost function that searches for the parameters that maximize the probability of observing the given data

#

it can be done in many different settings

brazen totem
#

shit if only I had an extra week to work on this

#

ah well I can always do it as like a curiosity after the due date

brazen totem
#

and for labeling I should only have one column for it right?

#

rather than 5 different columns with binary 0/1?

wooden sail
#

either is ok. the difference is whether you use categorical or sparse categorical cross entropy

brazen totem
#

what does that mean

wooden sail
#

the cost function to minimize is different, but the problems are equivalent

brazen totem
#

ah if they can have multiple classes at once?

#

ah nvm

#

they legit are just the same

wooden sail
#

yep

brazen totem
#

im lazy so id rather have 1 column only

full oar
#

Hey everyone do anyone know how to implement this kind node chart in python in gui

mighty spoke
#

Hi i'm trying to create an algorithm to bin my data but i'm not sure how to build it, like i could create a for loop and an if statement for each but that would be very long

unknown.png
versed gulch
#

Does anyone know if Python has any deconvolution algorithms for image preprocessing/processing and if so could you share the links to them. As I've only found the weiner and richard_lucy ones?

eager cloak
#

Heya!

#

How can I make it so that it just prints the response? (in this case, Hi, my friend, what can I do for you today?)

unknown.png
wooden sail
# mighty spoke Hi i'm trying to create an algorithm to bin my data but i'm not sure how to buil...

if you already have the values of t in a vector and have chosen T, you can compute all the values of tau. then, once you have the tau values, apply to all of them the procedure described in the text. in reality, splitting tau into bins follows the same formula that one uses to find the "phase" of the observations. having chosen M, you divide the period T by M, let's call this w = T/M. this is the width of each bin. then to find the bin index, you take tau/w and round down, i.e. floor(tau/w). if you have the values of t and/or tau in a vector, this requires no loops at all if using numpy arrays, for example.

cursive quest
eager cloak
#

❤️ kiss_void

cursive quest
mighty spoke
# wooden sail if you already have the values of t in a vector and have chosen T, you can compu...

Hi so I have the values of tau(phases) in a array and I tried using pandas to bin it but i'm not sure its working right

Tp=1/fr#time period
print(Tp)
phases = foldAt(t, Tp, T0=0)

plt.figure()
#data is the intensity


x7, y7 = zip(*sorted(zip(phases, data)))#ensures x and y values correspond to each others in pairs when sorted

plt.plot(x7, y7)

df4 = pd.DataFrame({'X' : x7, 'Y' : y7})  #we build a dataframe from the data
M=20#no of bins

bins=np.arange(0, max(phases), step=Tp/M)


categorical_object = pd.cut(x7, bins)
count=pd.value_counts(categorical_object)
grp = df4.groupby(by = categorical_object)        #we group the data by the cut
ret = grp.aggregate(np.mean)```
wooden sail
#

pd.cut looks likei t should do what you ask it to, yeah

#

pd is what you imported pandas as, right? should it be count = x7.value_counts(...) ?

native ibex
#

I am new to data science and ai and I am a self taught learn.
What are basic topic to learn

wooden sail
#

you wanna check out stuff like numpy, pandas, tensorflow, pytorch, jax on the python side

#

and at least some multivariable calculus, linear algebra, statistics, and optimization on the math side

mighty spoke
mighty spoke
wooden sail
#

seems about right. it'll depend on your actual data though, you'd have to plot it and see

#

try df4.value_counts, then? i'm not sure what a good way to do this with pandas is, tbh. i'd just do it with numpy

mighty spoke
mighty spoke
wooden sail
#

just do the math operations on the phases. or easier, use numpy histogram, since it seems that's what you're trying to make

mighty spoke
wooden sail
#

this would do everything for you

woven coral
#

How to use TF IDF vectorizer with GRU in Keras Python

#

anyone knows???

serene scaffold
woven coral
#

fake news detection

#

??

odd meteor
arctic wedgeBOT
#
Resources

The Resources page on our website contains a list of hand-selected learning resources that we regularly recommend to both beginners and experts.

odd meteor
#

Regression, Classification, Dimensionality Reduction (LDA, PCA), Clustering. Those are the basic topics

odd meteor
# woven coral How to use TF IDF vectorizer with GRU in Keras Python

You'd need to feed your tokenized text to your NN. So apply TF-IDF before feeding it to GRU. Or better still, use the tokenizer function in Keras directly.

from keras.preprocessing.text import Tokenizer
tokenizer = Tokenizer()
tokenizer.fit_on_texts('pass your input text here')

If you want to pad the sequence (text to sequence), then you need to also import the pad_sequence function from Keras

woven coral
#

thank u so much

#

why error????

odd meteor
woven coral
#

let me check

brazen totem
#

should I use density based hierchal or partional clustering

#

when I expect the different groups to have some overlap

woven coral
#

error

odd meteor
woven coral
#

vector

#

usig tf idf vectorizing

#

using

#

i think this method is not working on tf idf

#

its working on word2vec

frigid elk
#

what's a good resource for how to best structure my feature set for a binary classification problem of predicting an event 2 months in the future? .. not sure how my features should be built . ... one column for rolling 12 month average for sales vs 12 columns, one for every month in the past 12 months. ... or a mixture of both. ... looking for some examples

odd meteor
#

Replace where you have X with a random text, say, "I love cats" to see how it works. The problem is from your 'X'

frigid elk
#

obviously, other measures than sales, but looking for structure

odd meteor
median moat
#

What are Data Science projects that are good for a resume?

odd meteor
lapis sequoia
#

ayo what up anyone got any resources on cnn?

brazen totem
#

how do you determine cluster std for kmean

arctic cliff
#

So in normalization we should subtract the dataset by it's mean then divide by the standard deviation and here's a code

x -= x.mean(axis=0)
x /= x.std(axis=0)

where std from the documentation is simply applying the standard deviation formula with subtracts included:

std = sqrt(mean(x)), where x = abs(a - a.mean())**2.

The question is why should I subtract with the mean before dividing by the standard deviation?

brazen totem
#

you mean why can't you do it in one line?

arctic cliff
#

Nope
since the std formula already subtract by the mean
Why should I still subtract by the mean before dividing by the standard deviation?
Shouldn't I just do x /= x.std(axis=0) instead?

brazen totem
#

isn't x the value from the random sample?

arctic cliff
#

x is the sample

brazen totem
#

yeah

#

so you subtract mean to figure out how much this specific value differs from the mean
then divide it by the standard deviation to see how extreme that deviation is

arctic cliff
#

😳

#

It all makes sense now

brazen totem
#

yeah it's very logical

arctic cliff
#

Thanks a lot!

brazen totem
#

np lol

vagrant trench
#

hello guys , hope you doing well . pls I need help !

#

I need how to retrieve the number of clusters for the kmeans with elbow curve

serene scaffold
vagrant trench
#

yes

serene scaffold
#

did you use sklearn to do it?

vagrant trench
odd meteor
vagrant trench
#

another qst pls , how to plot the fitness curve vs Kmeans ?

odd meteor
# vagrant trench alright , I thought that it was another way to return it , thanks

There's actually another method to find the number of cluster(s) but Elbow plot is kinda more popular. You can get the number of clusters using KElbowVisualizer. You could do something like this

from sklearn.cluster import KMeans
from yellowbrick.cluster import KElbowVisualizer
print('Elbow Method to determine the number of clusters to be formed:')
Elbow_ = KElbowVisualizer(KMeans(), k=10, timings=False)
Elbow_.fit(train_df)
Elbow_.show()

If you don't fancy this method, you can use Silhouette plot as well.

What I usually do is, first use Elbow Plot, then validate the number of clusters using KElbowVisualizer.

odd meteor
vagrant trench
#

it makes error : No module named 'yellowbrick' !

odd meteor
orchid carbon
#

So.. I used this dataset I don't remember from where of weather pictures (literal sunshine, cloud, etc.)
What I did was use the KNN algorithim to classify the weather type. I am getting near 79% accuracy
The "features" I used (don't know exatcly how it works yet) was the medium of the R, G, and B proportions of every image pixels.
So all Red values of the pixels summed and divided by the amount of pixels.
Does this make sense?
It is analyzing the color context of the image, so as much as it ain't capable of identifying "clouds" or "sunrays" it can identify a color scheme of a "sunshine colored image" per say, I think

vagrant trench
vagrant trench
orchid carbon
#

:p

vagrant trench
#

any one here did a genetic algorithm for clustering ?

odd meteor
orchid carbon
#

Yeah. By NN you mean Neural Networks right

odd meteor
safe moss
#

Can anyone help, I have three dataframes that I want to perform difflib.get_close_matches on before merging them into one single dataframe but I am a bit stuck

vagrant trench
#

@odd meteor can u help me also this time pls 🥺 I want to plot time execution vs clusters number ( time execution in Y axe and number of clusters in X axe ) I don't know how

odd meteor
# vagrant trench <@519319496868233227> can u help me also this time pls 🥺 ... 
from sklearn.cluster import KMeans
num_clusters = range(1, 11)
inertia = []

for i in num_clusters:
    model = KMeans(n_clusters = i, random_state=42, init='k-means++')
    model.fit(train_df)
    inertia.append(model.inertia_)
    
plt.figure(figsize=(10,6))
plt.plot(n_clusters, inertia, marker='o')
plt.title('Elbow Plot: Number of Clusters vs. Inertia')
plt.ylabel('Inertia')
plt.xlabel('Number of Clusters (K)')
plt.show()
hoary wigeon
# 
max_temp = weather_df.groupby(['Station.State'])[['Station.City', 'Data.Temperature.Max Temp']]

I want to fetch the only city with max temperature

celest vine
#

Is it possible to scrape only those twitter profiles which have a certain profile pictures?
Suppose I want to scrape profiles with Bored Apes as there profile pic.

celest vine
hoary wigeon
#

then i want to find the city with max temperature in that state

celest vine
#

So, you want to find city with mac temperature for each state?

hoary wigeon
rose agate
brazen totem
#

how do you run dbscan machine learning with a ton of Nan values

#

I'm trying to use several columns in my prediction so naturally there's gonna be a lot of observations with at least 1 Nan

#

do I just restrict the model to columns where I expect Nan to be the lowest and drop em or is there a better way to deal with missing values

#

replacing everything with 0 will fuck up a lot of stuff

#

ill just restrict the model to columns where I expect Nan to be lowest and drop Nan for now but hopefully someone comes up with a better solution lol

#

what did I do wrong here lol

whatdidIdowrong.PNG
hoary wigeon
brazen totem
#

nvm figured out the graph

rose agate
safe moss
#

is anyone able to tell me how i would be able to keep the 0 index when i am filtering like this please?

unknown.png
#

i would like to keep all the questions intact

jagged hornet
#

I cant understand the error

err.png
jagged hornet
brazen totem
#

you have to explicitly single out the first row that's the only way I think

hoary wigeon
safe moss
brazen totem
#

nono OR

#

not and

safe moss
#

ah yes

#

i have tried OR and '|' but neither has worked

#

same result

#

wait maybe i need to reload the data

#

thank you that worked, with | rather than or

brazen totem
#

ah np

worthy phoenix
#

is there any way to update sklearn models to the newest version of sklearn?, so it works with the newer versions?

brazen totem
#

yeah you should use the -U

worthy phoenix
#

i mean i have the model dumped already, im just trying to load it with pickle and gives error

worthy phoenix
#

ModuleNotFoundError: No module named 'sklearn.feature_extraction.dict_vectorizer'

brazen totem
#

you tried uninstalling and reinstalling already?

worthy phoenix
#

i mean im on a separate work environment rn so ye i had to install everything from scratch

brazen totem
#

in what ive seen online it might just be misnamed

#

you checked the folder to make sure you're typing it right?

worthy phoenix
#

ye, thats not the issue, basically the issue is that when there is sklearn version mismatch. For example, trying to deserialize a sklearn(>= 0.22.X) object dumped with another sklearn version < 0.22.X. since sklearn introduced a change between those version

brazen totem
#

ahhh so you're trying to use old code with new functions

#

and it breaks down

worthy phoenix
#

yes

brazen totem
#

other than redoing the code the new way idk what fixes there are

worthy phoenix
#

pain, i dont have the dataset to train again

#

;-;

#

dont even remember how i found it

brazen totem
#

or maybe you can like manually install an old version of the packages

#

after uninstalling the new one

#

if you remember which version it was that is

worthy phoenix
#

ye it was 0.19.0

#

i remember that

#

but i need some dependencies that need higher versions of scikit-learn

#

so i cant even downgrade it

brazen totem
#

if this has any solution it's gonna be hella bullshit is all I know

worthy phoenix
#

ima google around more to see if there is a way

brazen totem
#

@wooden sail
you got any clue how to interpret this?

interpret.PNG
#

this is optics modeling of my poker data using 5 different variables normalized

worthy phoenix
brazen totem
#

LOL

#

yeah that's what I would expect

worthy phoenix
#

guess gotta find the training data again, rip, they should have maintained model persistancy for older versions as well

#

;-;

brazen totem
#

they dont care lmao

safe moss
#

can anyone tell me what the point of doing something like .fillna(0) to get rid of NaN is because as soon as I do something else to the dataframe the NaNs just come back?

worthy phoenix
#

ima just edit the pickle with the newer module names

brazen totem
#

hopefully that works ?

worthy phoenix
#

idk, thats the only hack which comes in my mind rn

#

anddddd it workeddddd

#

bruhhhhhhh

brazen totem
#

nice

odd meteor
safe moss
safe moss
#

is anyone able to assist with seaborn

sterile rivet
#

Any idea how to convert a 1gb json file to csv?

tidal bough
arctic wedgeBOT
frigid elk
sterile rivet
#

I have converted it using to_csv how do I download it?

sleek tapir
#

is anyone doing andrew ngs

#

course (again next month)

rotund ledge
#

I'm trying to get an animation of filling in a 3d line (to visually represent a line integral)
Effectively this: https://pythonguides.com/matplotlib-fill_between/#Matplotlib_fill_between_animation
It's not working though

import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import matplotlib.animation as mtA
import numpy as np
from mpl_toolkits.mplot3d.art3d import Poly3DCollection

nsteps = 10000
t_init = 0.0
t_final = 20.0
t=np.linspace(t_init,t_final, nsteps)

rectangles=30
val = np.empty(rectangles + 1)
ts = np.empty(rectangles + 1)
tp = np.empty(rectangles + 1)

#parameterization of curve
def x_param ():
    #x=np.cos(t)
    x=t
    return x

def y_param ():
    #y = np.sqrt(t)
    y=t
    return y

def z_param ():
    z=t
    return z

def reinman ():
    pass

def main ():
    fig =plt.figure()
    ax=fig.add_subplot(projection='3d')
    #line, = ax.plot([], [], [])
    ax.plot(x_param(),y_param(),z_param(),color="orange")

    set01xtoline = [x_param(), y_param(), np.zeros(len(t))]
    set1 = [x_param(), y_param(), z_param()]
    zZ=np.zeros(len(t))

    def animate(i):
        x3=x_param()
        y3=y_param()
        z3=z_param()
        xA=x3[0:(i-1)]
        yA=y3[0:(i-1)]
        zA=z3[0:(i-1)]
       
        p = fill_between_3d(ax, xA,yA,zZ, xA,yA,zA, mode=1, c="C0")
        return p

    ani = mtA.FuncAnimation(fig,animate, frames=1000, interval=50000)
    #fill_between_3d(ax, *set01xtoline, *set1, mode=1, c="C0")
    plt.show()

if __name__ == '__main__':
    main()```
#

Fill 3d function I'm using

def fill_between_3d(ax, x1, y1, z1, x2, y2, z2, mode=1, c='steelblue', alpha=0.4):
    if mode == 1:

        for i in range(len(x1) - 1):
            verts = [(x1[i], y1[i], z1[i]), (x1[i + 1], y1[i + 1], z1[i + 1])] + \
                    [(x2[i + 1], y2[i + 1], z2[i + 1]), (x2[i], y2[i], z2[i])]

            ax.add_collection3d(Poly3DCollection([verts],
                                                 alpha=alpha,
                                                 linewidths=0,
                                                 color=c))

    if mode == 2:
        verts = [(x1[i], y1[i], z1[i]) for i in range(len(x1))] + \
                [(x2[i], y2[i], z2[i]) for i in range(len(x2))]

        ax.add_collection3d(Poly3DCollection([verts], alpha=alpha, color=c))```
tacit basin
vagrant trench
#

hello guys , I want to plot time execution ( in y label ) vs clusters number ( in x label ) , how can I do it pls ? PS : I have the time execution and clusters number

distant horizon
misty flint
#

theres 4 challenges

#

one of them is audio analysis

#

think i want to try that one