#serious-discussion

1 messages · Page 108 of 1

pulsar pagoda
#

the problem la the validation

peak tide
#

you're not, you'd be reinventing the wheel

velvet dagger
#

ML/DS... there's the research which feels kinda statistical in nature

#

And then there's doing stuff in practice which basically seems to be

#

from sklearn.something import whatineed

loud snow
#

the research seems cool and creative

jagged forge
#

if you like data structures and algorithms, go into database engineering

pulsar pagoda
#

thats cause your doing python

jagged forge
#

or big data systems in general

velvet dagger
#

And in practice the decision making process is closer to

peak tide
#

ML is the most mathematically interesting side of CS, at least that i've dealt with

pulsar pagoda
peak tide
#

although some of the numerical math stuff can be interesting too

loud snow
#

yeah

jagged forge
pulsar pagoda
#

i Guess You could be a horrible architect

velvet dagger
#

Which tool do I use? Do some EDA, is this supervised or not? Classification/regression? Which algo works best? etc

loud snow
#

the numerical stuff is pretty cool

pulsar pagoda
velvet dagger
loud snow
#

nah it isnt

pulsar pagoda
loud snow
#

using python is industry standard for a lot of ml things

pulsar pagoda
#

if ur doing a prototype

#

no

loud snow
#

it is lol

peak tide
#

using python is industry standard for a lot of things, period

jagged forge
# pulsar pagoda wouldnt cloud make this easier?

if you're saying big data is easier than databases, consider that big data systems need to be able to scale massively, and the problems of distributed computing only get worse the more machines you add

loud snow
#

python is in excel now too

#

so fuck VBA

peak tide
#

ha excellent

loud snow
#

will be in excel*

#

no need for pandas either

#

all of my data will always be clean how I like it

jagged forge
#

also big data and databases aren't really mutually exclusive, i guess i'm making it sound like that

pulsar pagoda
#

cloud makes it easier

jagged forge
#

what knowledge and tooling are you referring to

loud snow
#

what do you mean?

pulsar pagoda
#

for example

loud snow
#

if i think i get what you mean

#

a lot of people who own large data bases

#

dont know shit about their machines

#

people meaning companies

pulsar pagoda
#

You can't make a single instance of python handle a large data size in one proceds

jagged forge
#

yeah if you're just using the products that process the big data then sure

loud snow
#

not a lot of companies are good at keeping inventory

pulsar pagoda
#

so u can just paralelize with containers ur workload

loud snow
#

and this furhter goes down to databases

jagged forge
pulsar pagoda
#

yeah

jagged forge
#

in case you didn't realize, most databases are distributed

pulsar pagoda
#

but in c++ You would use threads and have more control

#

for example

river moon
#

there was that one yahoo problem a couple of years ago where they ran a contest on who can figure out a way to efficiently solve a logistic regression problem with like 10^6 or 10^7 samplesbleak

pulsar pagoda
#

like i'm not saying it's not a problem

loud snow
pulsar pagoda
#

but the average is not going yo have the knowledge for handling it

loud snow
#

one thing ive learned is that python is abusrdly general now

jagged forge
#

i'm not talking about the average person

pulsar pagoda
#

it doesnt have to be c++

loud snow
#

so whatever feature you think it doesnt have, someone probably wrote a library with

pulsar pagoda
#

it could be c#

jagged forge
#

i am talking about the people building big data systems and infrastructure and the people building databases

#

most people do not fall into that category

#

i am not talking about a python dev that builds data pipelines on top of that infra

pulsar pagoda
loud snow
#

what is in the edge?

pulsar pagoda
#

resource constrained environments

#

iot

#

for example

jagged forge
#

it all ends up in the cloud anyway, otherwise you're not talking about big data

pulsar pagoda
#

nooe

#

nope

jagged forge
#

if it didn't need to go in the cloud, it's not enough data to be big

loud snow
pulsar pagoda
#

Big data is when ur data is bigger than the memory u have

loud snow
#

people have big data centers

jagged forge
#

no.......

pulsar pagoda
#

thats the formal definitiion

jagged forge
#

big data is when your data is bigger than a single machine or even a few machines can hold

#

that's cap

pulsar pagoda
#

definition*

jagged forge
#

there's no formal definition of big data, but if there was, that ain't it

pulsar pagoda
#

by memory i mean computer memory

#

as in ram

jagged forge
#

yeah and that's blatantly wrong

#

you don't have big data just because your data needs to be stored on disk

#

you have big data if you need a distributed file system or object store to manage your data because it's so massive

#

we're talking at least tens of terabytes

#

and that's not even that much

#

petabytes is common

peak tide
#

that would be my understanding as well.. if you can access it via a simple file stream it ain't big data

loud snow
#

maybe a good definition is if you need to consider a truck to transport data as a viable means of transporting data

#

then its big data

jagged forge
#

LMAo they literally have that you know

#

it's called snowmobile i think

loud snow
#

i know lol

jagged forge
#

my company had to get data delivered in one

loud snow
#

my company does it

jagged forge
#

big data problems....

loud snow
#

the most important lesson i learned doing my job

#

is that i should start a tech company if I want to be fuck you rich really quick

jagged forge
#

well

loud snow
#

there is so much money available

pulsar pagoda
#

hire me

#

lol

jagged forge
#

if you're willing to have max levels of stress in the beginning maybe

loud snow
#

yeah idk

jagged forge
#

and if you think you actually have a shot

loud snow
#

i am planning on starting a company when im into late 20s

#

so that I have experience and fail safes

jagged forge
#

our ceo told us this story one time where he was locked out of the data center during an outage and he slept outside in nyc in the cold on a bench

#

for like 2 hours

pulsar pagoda
#

u know that the average salary for a dev here is less than 2k right?

loud snow
#

tough life lol!

#

where are you?

pulsar pagoda
#

not kidding

loud snow
#

costa rica?

jagged forge
#

is that at least high in terms of your cost of living

loud snow
#

2k USD?

pulsar pagoda
#

yes

#

costa rica

loud snow
#

2k usd per year?

#

thats a lie

pulsar pagoda
#

no

#

monthlt

#

lol

loud snow
#

bruh

#

thats fine then

pulsar pagoda
#

monthlt*

#

fuck

sly flint
#

That's pretty okay.

pulsar pagoda
#

autocomplete

jagged forge
#

apparently in most other countries they post monthly instead of annual salaries

loud snow
#

i dont know much but im assuming houses are in the 150-250 range

jagged forge
#

(other as in not the usa)

#

im racist sorry

pulsar pagoda
#

cheaper

loud snow
#

bruh

#

then shut up tbh

pulsar pagoda
#

we have 5 100k houses

loud snow
#

i can't even dream of getting a house xd

pulsar pagoda
#

at least My mom

#

has

loud snow
#

2k is ridiculous then

peak tide
#

you might well be better off than someone in silicon valley making $250k/year

loud snow
#

not joking you are

pulsar pagoda
#

health is cheap too

jagged forge
#

my hope is to continue doing remote work and then buy a home somewhere that's not ridiculously expensive

peak tide
pulsar pagoda
loud snow
#

one of my friends does her job and rock climbs while doing it

jagged forge
#

but i also hate driving..... and nyc is walkable..... also it's the capital of finance.....

#

😭 decisions

pulsar pagoda
#

but if u know ur shit u can live with no trouble

loud snow
jagged forge
loud snow
#

ok makes sense

peak tide
loud snow
#

im from nyc

#

nah bungo

#

im not gonna lie

#

every city is shit with public transport id say

#

argueably not boston or dc

peak tide
#

sure, but why do you need public transport if you're working from home

loud snow
#

but every city ive been to sucks a lot

jagged forge
loud snow
#

freedom

pulsar pagoda
#

i think nyc might be bigger than costa rica

loud snow
#

you can't go anywhere without a car in greater america

jagged forge
#

the next place i live will be cold, because fuck hot and humid places

#

fuck houston

peak tide
loud snow
#

):

#

such a sad cope i hear from common nyc haters

#

rent is expensive yes

peak tide
#

i don't hate nyc, i'm pretty neutral on it

loud snow
#

but there are affordable places

peak tide
#

if i had chosen to work in finance instead of tech, i'd have moved there of course

jagged forge
#

right, i would be closer to the finance industry

peak tide
#

but that was before the paradigm shift of being able to work from anywhere

pulsar pagoda
#

now that u mention it

loud snow
#

better, closer to me

pulsar pagoda
#

arent there data types that are super common in finance that have to do with precisión of decimal digital?

peak tide
velvet dagger
#

Being from NY is morally correct but life there is tough

pulsar pagoda
#

digits*

jagged forge
velvet dagger
#

Bay Area does win big time on climate

loud snow
#

oh ye

#

suremark

pulsar pagoda
loud snow
#

werent u asking about SBE decoding from CME or something a while ago

jagged forge
#

im so jealous

loud snow
#

or parsing SBE

jagged forge
#

don't remind me of that shit

pulsar pagoda
#

who?

peak tide
loud snow
#

if i dont do something similar in a year im going to do it in 2 ):

peak tide
#

i lived in newport beach for a while, that was pretty nice

jagged forge
#

fuck cme and their obtuse data feeds and their xml schemas and their garbage code generators and their overengineered stack

#

and their FTP server that kept rate limiting me when i accessed it more than 10 times in an hour

#

(idk what the actual rate limit was)

loud snow
pulsar pagoda
#

they actually host their own ftp?

#

really?

jagged forge
#

yeah

pulsar pagoda
#

cancer

loud snow
pulsar pagoda
#

the fucking management overhead

#

like put it on the cloud

jagged forge
#

actually im not really even sure if you can call it an ftp server

pulsar pagoda
#

idk

loud snow
#

im sure they have to do it for security reasons

jagged forge
#

so it's actually just a web server

pulsar pagoda
#

yeah i remember when i had to follow hippa

#

and My employer was like

jagged forge
#

anyway what i ended up doing is not logging in, because apparently they used my account identity to rate limit me

loud snow
#

if it hosts ftp connections it can still be ftp server right?

pulsar pagoda
#

break hipaa

jagged forge
#

and i browsed their ftp website with the blur on

pulsar pagoda
#

cause i need it

loud snow
#

but hippa violations are expensive

pulsar pagoda
#

i rather follow the rules if uncle Ben is on top

loud snow
#

so if you get caught then rip

pulsar pagoda
#

sam*

#

uncle sam

loud snow
#

i didnt know costa rica had hippa

pulsar pagoda
#

no

#

but the client was in the us

loud snow
#

o

#

yeah u gotta follow parent country rules ig

#

wouldn't it be illegal if they sent patient information to your company regardless?

pulsar pagoda
#

like mf wanted me to scp files from his servers to My pc

loud snow
#

idk who audits these things tbh

#

so you would probably be fine

pulsar pagoda
#

and bro had terrible connection

loud snow
#

not them tho

pulsar pagoda
#

so forget vscode

#

ssh and vim

jagged forge
#

you can do vscode over ssh tho

pulsar pagoda
#

yes

jagged forge
#

oh

#

but the internet was shit

pulsar pagoda
#

but the latency is terrible

jagged forge
#

i see

pulsar pagoda
#

lol

loud snow
#

i know the feel

#

it sucks so much

pulsar pagoda
#

worst part is most people wouldnt ssh and set up their vim for work

#

cause u know

#

pepegs

loud snow
#

you like calling people pepegs

pulsar pagoda
#

yeah that too

loud snow
#

i understand

velvet dagger
#

Not even sure what it means haha

jagged forge
#

pepegs is spanish for pomegranate

loud snow
#

i think they mean the twitch emote

pulsar pagoda
#

indeed

jagged forge
#

it's actually a term of endearment

loud snow
#

suremark you read jjk?

jagged forge
#

no but i have been following the manga loosely

pulsar pagoda
#

jjk?

jagged forge
#

and i heard about the latest chapter

loud snow
#

well i want you to get caught up

#

yeah it was cool

#

the entire fight

jagged forge
#

i liked jjk the anime but i read the manga and wasn't really a fan tbh

pulsar pagoda
#

ahh

#

jujutsu kaisen

#

yeah

loud snow
#

yeah its an aquired taste

pulsar pagoda
#

get caught app

jagged forge
#

just too much action not enough uhhhh plot i guess

loud snow
#

):

pulsar pagoda
loud snow
#

not truuu

#

like black coffee

jagged forge
#

yeah idk my brother thinks it has good plot

#

i just wasn't convinced by it ig

pulsar pagoda
#

i had to quit black coffe after My last job

loud snow
#

best plot manga rn?

pulsar pagoda
#

the stress caused acid reflux to develop

loud snow
#

im consistently happy with the fable

#

9 chapters came out

loud snow
jagged forge
#

well the only mangas i'm reading rn are one piece, opm, and chainsaw man

#

cuz im ultra normie

velvet dagger
#

I'm not really team coffee anyway

#

Coffee tastes like death

pulsar pagoda
jagged forge
#

well, i happen to like death if you mix it with enough creamer and sweetener i guess

pulsar pagoda
#

but nahh

#

it was a 1 man job

#

only dev of the company

loud snow
#

):

#

I wish I was the only dev of the company

jagged forge
#

maybe you dont

loud snow
#

because I would completely take advantage of management

jagged forge
#

i heard it's not actually a very good thing

pulsar pagoda
#

nope

velvet dagger
#

Yeah I mean something like caramel latte is fine

loud snow
#

i know two people who are only devs

jagged forge
#

i am glad i work at a company where the product is software and most of the people are software engineers

loud snow
#

and people shit on them

pulsar pagoda
#

cause they expect u to do everything

loud snow
#

id rather be shit on and irreplacable then replaceable and praised

loud snow
pulsar pagoda
#

i was shit on and irreplacable

jagged forge
#

you can be irreplaceable and praised

loud snow
#

i feel like I shouldnt be so harsh sometimes

pulsar pagoda
#

and they still fired me

jagged forge
#

just become the maintainer of a legacy software system

velvet dagger
loud snow
velvet dagger
#

Should be my motto tbh

pulsar pagoda
loud snow
#

if you are shit on enough you know that your work isnt valued

#

so you have to leverage this to the tits

pulsar pagoda
#

there was also the fact i told the client no

#

so

velvet dagger
#

I mean... I think being irreplaceable in something like tech is probably overvalued

#

People replace themselves to get ahead

jagged forge
#

true

loud snow
#

the situations im talking about you usually are remote

#

so you can pretty much collect free money

jagged forge
#

but i will say i would not have had the chance to take on my current, very ambitious project if my boss considered me replaceable

loud snow
#

and write trash infrastructure

jagged forge
#

i would have been fired when that recruiter sent my resume to my boss

loud snow
#

wasting company dollars legally

pulsar pagoda
loud snow
#

nah

jagged forge
#

yaml engineering hype

loud snow
#

software industry doesnt usually hire 1 software person

pulsar pagoda
jagged forge
#

fuck that shit

loud snow
#

yeah

#

my friend telling me about it

pulsar pagoda
#

it's not

#

what's complicated

loud snow
#

he really went thru it

pulsar pagoda
#

is their pricing

jagged forge
#

have you ever tried to set up a lambda

pulsar pagoda
#

yes

neat lintel
#

It's black coffee or its nothing

pulsar pagoda
#

i have

loud snow
#

nah im pretty sure its complicated networking wise

#

among other things

pulsar pagoda
#

with SAM

#

and serverless framework

jagged forge
#

i did and i just went back to digitalocean cuz it was way simpler and i was just doing a tiny project

#

also their auth is a whole thing

pulsar pagoda
viral rivet
#

hello good people

#

i need your guys' opinion on what i should do

pulsar pagoda
#

sure

#

talk

viral rivet
#

i went to a job interview and the manger never said i was hired, but to wait until he does other interviews cause he has a few left to do

#

but he sent my info to the owner and i completed onboarding

pulsar pagoda
#

did he sent u an offset?

#

offer?*

loud snow
#

do a followup

viral rivet
#

and the last step was to contact the manager to get my schedule, so i did and he hasnt replied 😼

loud snow
#

just follow up

#

people get swamped with emails and have bad management systems

pulsar pagoda
#

yeah do what cofe says

viral rivet
#

he doesnt have email

loud snow
#

do you want to work there then?

viral rivet
#

i texted him about the schedule days ago and he hasn't said anything

#

YES I DO

loud snow
#

fuck that lol

viral rivet
#

I NEED THE MONEY

loud snow
#

he doesnt have email

#

you srsly dont have pride?

viral rivet
#

no

#

i am so desperate

loud snow
#

please dont be

pulsar pagoda
#

i get him

viral rivet
#

i've applied to 10 places and this is the first to give me an interview

loud snow
#

its not worth working with someone without emial

#

are you in states?

viral rivet
#

it's domino's so it's fine

loud snow
#

lol

pulsar pagoda
#

ahh

viral rivet
#

yeah im in the US

loud snow
#

nvm

#

do whatever tbh

#

just show up even

viral rivet
#

if i were to text him what should i say tho?

loud snow
#

whatever you do, dont take the job seriously

viral rivet
#

i cant think of anything

loud snow
#

well text him what you told us

viral rivet
loud snow
#

politerly or something

#

i dont mean it in a bad way

pulsar pagoda
#

yeah

viral rivet
#

i think i have an idea of what to say but idk how to open it

loud snow
#

but i hate you if you work at dominoes

#

ive had 3 hour quality checks before

viral rivet
#

💀

#

dominos wasn't my first pick

#

but it's the best ive got

#

then it's taco bell 😭

pulsar pagoda
#

it's a pick alright

#

and it's more than i have

viral rivet
#

he said he would give me $13 an hour after my first week

#

to make pizzas

pulsar pagoda
viral rivet
#

thank you all 🙏

neat lintel
gusty canyon
noble osprey
#

Hey guys, Gangster Mathematician here 🤙

neat lintel
#

Def

neat lintel
noble osprey
#

and mathematician

neat lintel
noble osprey
#

you no is who I am cuhh

burnt ledge
#

💀

neat lintel
noble osprey
neat lintel
#

Stop joking lil bruv

noble osprey
#

Im a gangster mathematician

neat lintel
#

Ain't no gangsta goin hop on discord tellin folks he a gangsta u get me sotrue

gusty canyon
#

may I have permission to post attachments in here?

neat lintel
noble osprey
#

ion care im a gangster mathematician

burnt ledge
gusty canyon
#

Ok

noble osprey
#

cuz I smoke em both either way

surreal sapphire
#

this channel is for serious discussions, please

solar hawk
vapid sentinel
#

Principles and Techniques in Combinatorics" by Chen Chuan-Chong and Koh Khee-Meng
is this book good enough for learning combinatorics from scratch?

little vine
#

I'm so frustrated with math omg my day is ruined

solar hawk
#

Cali why :(

little vine
#

I'm stuck on a problem

#

I threw the kitchen sink at it

#

Then looked at a solution but don't really get how it solves it

#

So Im taking a break and looking at again

#

Feeling like a charlatan lolopencry

little vine
#

Return to math

#

Never return to math

#

I'm just going to do a different problem

flint basin
long matrix
#

ill do what i want with the sink.

noble osprey
#

Sigh gangster mathematician hereemoti_sad

tame egret
#

hihihi!

ocean harbor
#

helohelo

zinc lily
#

I like to find problems for the existing solutions.

tame egret
#

wdym

prisma lake
#

Hello

#

Much more quieter and pleasant here

wintry wadi
#

time to rise and shine

prisma lake
wintry wadi
#

it's the night

#

but thanks lol

prisma lake
#

I came in the chat area to discuss and get thoughts on how should I approach my studies in math, but holy shit the other discussion chat is crazy

prisma lake
wintry wadi
#

not to be rude or anything but

#

do we need to tell you that 😭

#

i mean your in the server for the same reason most of us are here in the server

#

you seem pretty interested already

#

probably approachin your studies in math in a good way

#

maybe im misreading

tame egret
wintry wadi
#

ok it seems like i misread

wintry wadi
tame egret
prisma lake
wintry wadi
#

ah ic

prisma lake
#

Also the reason why I joined this server.

wintry wadi
#

well joinin here is a start

#

main thing is dont slouch

prisma lake
wintry wadi
#

past papers help

#

r u pre uni too?

prisma lake
#

I am working for the most part. I do want to get back in college and strive for a degree.

wintry wadi
#

yeaah past papers help then

#

best thing is just ask

#

tbh the help im giving u here could be applied to like any other subject lol

prisma lake
prisma lake
wintry wadi
prisma lake
#

I don't have them anymore.

#

I moved and during moving I got rid of them.

honest storm
#

Is there a generic name for the property of a set of objects that are isomorphic to a canonical object?

#

Like for example say the canonical object is the real line

#

so the set is

#

${x: x \approx \mathbb{R}}$

fathom swallowBOT
#

Makogan

220634889576382464.png
honest storm
#

I wonder if there is a single word

#

Like instead of saying "let us cosnider the set of all objects isomorphic to blah" I want to say something like "The set of all blah ismomorphables" or soemthing like that.

storm sage
#

or more generally, equivalence class of x

honest storm
#

Thank you

storm sage
#

note that such a thing is probably not a set though

#

because then you could make funky stuff like the set of all sets

#

but saying the "class" or "collection" of stuff isomorphic to x is pretty common

versed nimbus
#

Its official, Im taking ap calc for 11th grade next year!

little vine
#

Grats homie

versed nimbus
neat lintel
#

" (I'll also leave as an exercise for you to define directed graphs set-theoretically! Hint: E ⊆ V ×V, but you may need to take bags into account.) " despair

storm sage
#

or, "defined to be"

pulsar pagoda
#

@jagged forge i can continue this conversation later

#

i gotta go out

#

for a bit

#

will ping u when im b

#

k?

jagged forge
#

ok catshrug

solar hawk
#

Oh it’s C

pulsar pagoda
solar hawk
long matrix
#

wait how do these packages work

#

in a non-cpython

solar hawk
#

Idk

#

I think it might be like Node

pulsar pagoda
solar hawk
#

With Nodes c Interop thing

pulsar pagoda
#

it has a runtime thats built upon

#

making things Even worse for python

#

i'm about to go out just waiting for my brother

pulsar pagoda
#

not a c one

#

as far as i know

solar hawk
#

Yeah I was thinking of that

jagged forge
#

i.e. most of the important stuff

#

i would find it more plausible that numpy drags behind octave in hpc benchmarks because parallelism in python sucks unless the c/c++ library you're using does it for you

solar hawk
#

C++

pulsar pagoda
#

im back

#

also the c++ math libraries might be better written than the c ones

#

consider numpy uses c inn core/src

loud snow
#

wat

jagged forge
#

"better written" is a weird way to put it, maybe you mean it benefits more from templates and things of that nature

loud snow
#

also

#

you see the black bar

#

guess the language

pulsar pagoda
#

c

pulsar pagoda
#

templates are resolved before execution

jagged forge
#

yeah yeah inlining and all of that, i know

pulsar pagoda
#

like bro

#

im not saying dont use python

#

im saying know your tools

#

there is also a python runtime which is made for scientific computing

#

cant remember the name right now

jagged forge
#

you made a blanket statement about numpy being slow and i cast doubt on that

#

that's all

#

in certain contexts python can be on par with c/c++

#

(because the libraries are written in c/c++)

jagged forge
weary sable
#

hello all

jagged forge
#

well, "not much" is debatable, but you don't get as much

pulsar pagoda
pulsar pagoda
jagged forge
#

when you're doing processing of that nature, it doesn't matter too much if you have to do a few allocations or a few ffi calls into C code from python, since it's a few times per batch

#

i am pretty sure any dynamic language with efficient array operations works the same way

pulsar pagoda
jagged forge
pulsar pagoda
#

yes but if your coding for a device thats supposed to deploy to a field in suboptimal conditions (no constant access to electricity, hard access to a network). What do u use?

peak tide
jagged forge
pulsar pagoda
#

you choose your tool depending on your problem

jagged forge
#

i don't disagree with that, i disagree with python automatically being slow

pulsar pagoda
#

now if im being paid and can use the cloud

#

thats a different story

jagged forge
#

i actually thought python is common in embedded development

#

though not more common than c/c++

#

but like, just as a control plane kind of thing

pulsar pagoda
#

like meson is actually really ez to learn (build system)

#

plus if your trying to set up scripts for handling your environment

jagged forge
#

so when we say embedded, are you talking about like, raspberry pis, or are you talking about running code on bare metal

pulsar pagoda
#

running code on bare metal

#

with hard real time constraints

#

where you have to use RTOS

jagged forge
#

yeah that stuff sounds awesome, but i hear it does not pay well sadly

pulsar pagoda
#

and u cant have a scheduler

jagged forge
#

otherwise i would be really interested

pulsar pagoda
#

like

#

i just want a job where im not gonna be busting my balls trying to justify my technicals decisions

#

to people who dont even know what a compiler is u know

#

that stems from me

#

being terrible at dealing with people

jagged forge
#

my company is very software oriented, so i have never really ran into those issues

#

we basically build a data platform for other software developers

#

so everything is oriented around software

pulsar pagoda
#

well you also have the advantage of living in the states

jagged forge
#

software isn't as big in costa rica, i guess? or...

pulsar pagoda
peak tide
#

there's embedded SW development at most of the FAANGs these days, they do pay pretty well

pulsar pagoda
#

that dont like math

jagged forge
pulsar pagoda
#

they think ML

#

is calling API's from aws ml services

#

instead of developing ur own models

#

thats how bad it is

peak tide
#

which is developed by a bunch of smaller teams that were mostly acquired by the big guys at some point or another

pulsar pagoda
jagged forge
#

my boss was interested in them but we couldn't get one that fit in our server racks

pulsar pagoda
#

@jagged forge the market is moving towards ML accelerators

peak tide
pulsar pagoda
#

thats where the major developments are being made in architecture

#

as far as i know

peak tide
pulsar pagoda
#

which is a big topic

#

and fpga's actually make sense for that (i think they do i have to ask a electronic/eletrical engineer).

#

cause u can prototype and simulate the components beforehand

#

so less testing on assembled final products

peak tide
#

fpga's are used before the actual chips are ready, main disadvantage is that they're usually much slower than the real chips

pulsar pagoda
#

@peak tide or they consume more energy

peak tide
#

also sometimes they use jank-ass interfaces that differ from the real ones, so you have to write one-off special sw just to talk to them

pulsar pagoda
#

by interfaces u mean software?

#

or hardware interfaces?

peak tide
#

(not in smartphones though!)

peak tide
pulsar pagoda
jagged forge
pulsar pagoda
pulsar pagoda
#

guess where you prototype those?

peak tide
pulsar pagoda
jagged forge
pulsar pagoda
#

i do c++

#

so HLS (high level synthesis)

jagged forge
#

what does that stand for

#

google's not helping me out

#

oh, high-level synthesis?

pulsar pagoda
#

yes

#

for our publication

#

but turns out Xilinx claims its a configurable fft but its not

#

xD

#

at this point i would call my self more of a enthusiast though

#

my goal is to be proficient by the time im a grad student

peak tide
#

gonna guess that fft does not mean fast fourier transform in this context?

pulsar pagoda
#

it means fast fourier transform

peak tide
#

oh..

#

what do you do with it?

pulsar pagoda
pulsar pagoda
#

he is targeting accelerators for ml models that use quantize dense, and convolutional layers

peak tide
#

oh nice

#

yea ffts can be used to do convolutions much faster

pulsar pagoda
#

like ive been learning this stuff for over the past 4 months

#

i can sort of get where some of the pieces fit but its not like i would say im at a "senior" level

#

but u have to start somewhere right?

peak tide
#

absolutely

pulsar pagoda
#

so let me give u more context bungo

#

he introduced me to this problem of quantization

#

and basically his idea for my career development is to have a thesis on quantization for fixed point notation

#

ive read a bunch of papers of the state of the art

#

found out my major doesnt teach me nearly enough maths for it

#

so i study the math in my spare tie

#

time*

peak tide
#

quantization, like using some kind of compressed format to represent fixed point numbers, or what does it mean

#

oh this is a CNN thing apparently, i'm googling a bit

pulsar pagoda
#

its for deep learning

#

when he introduced me he passed me a link

peak tide
#

yea i see

pulsar pagoda
#

let me send it to u

peak tide
#

oh i see you train the model to learn efficient quantized representations

pulsar pagoda
#

yes

#

but there is also the problem of not having enough data

peak tide
#

yea always haha

pulsar pagoda
#

or federized learning

#

i think its called

#

federated*

peak tide
#

yea

#

like crowdsourced learning if you deploy it on phones or whatever

pulsar pagoda
#

when we started he started talking about numerical methods, he suggested stochastic gradient descent

#

and im like

#

nope

#

definitely not enough math

peak tide
#

yea that's like the fundamental algorithm for many learning methods

#

trying to search the parameter space in some reasonably sensible way to find a local minimum

#

you're never gonna find the global minimum haha

pulsar pagoda
#

since u know you use supremas and infimas all the time

peak tide
#

that's part of the theory behind the scenes

pulsar pagoda
#

like i do like math

peak tide
#

maybe invoke some functional analysis to prove that the algorithm actually converges

pulsar pagoda
#

but if people that know more math than u tell u its overkill you start to think u know

peak tide
#

i think for stochastic gradient descent in particular you just need multivariable calculus and probability/statistics to understand what it's doing

#

in fact that's all you need for most machine learning (well that and linear algebra)

pulsar pagoda
peak tide
#

i'm no expert, but most treatments of ML that i've read were fairly strongly bayesian

pulsar pagoda
#

well i guess i at least delimited a little the domain of expertise with this conversation

#

so within which topics of mathematics would this be categorized?

peak tide
#

well the CNN stuff in particular brings in convolutions and thereby fft's

#

so understanding that stuff is nice

#

fourier analysis basically

pulsar pagoda
#

applied fourier?

peak tide
#

yea

pulsar pagoda
#

u dont need measure theory or any of that stuff?

peak tide
#

no, not at all

#

you're doing discrete operations and even if you view them as approximations to continuous operations, you're not dealing with data that's so pathological that it would make a difference to use riemann vs lebesgue integration

#

and you're not doing anything so sophisticated with probability that you need to worry about the measure-theoretic underpinnings

#

although that stuff is cool and it's worth learning if you're interested

pulsar pagoda
#

for now the answer is more math but more straight forward than i thought and with a clearer limit on topics covered

#

which is fine for now

peak tide
#

yea sounds right

#

if you want to understand CNNs in particular those are the topics i would look at

#

maybe read a bit on signal processing from the engineering viewpoint as well

#

to get the nitty gritty about how the fft algorithm works

#

(also covered in some CS algo books)

pulsar pagoda
#

i have a book on the engineering view point

peak tide
#

cool, for a more mathematical treatment but still more application-focused vs. theoretical you could try folland's book on fourier analysis

#

for a more theoretical approach but without requiring measure theory/lebesgue integration, i always recommend stein and shakarchi vol 1, that's a really nice book

pulsar pagoda
#

right now im with sherbert

#

probably have to go through those when im done with this

peak tide
#

intro to real analysis or something similar?

pulsar pagoda
#

intro

peak tide
#

i first learned real analysis from another book by bartle actually (i think he's the co-author of that sherbert book)

pulsar pagoda
#

i have both volumes of apostol calculus too (physical)

#

was thinking on going through those for the linear algebra (he touches on that in those books)

#

i have done linear algebra in college

#

but it doesnt feel formal enough

#

also only gets to vector subspaces

#

and no proofs

peak tide
#

yea that should be fine, enough LA for your immediate purposes

#

apostol's kind of boring and dry to read but aside from that it has good coverage

pulsar pagoda
#

and then goes onto prove this

#

and im like

#

wtf

#

im supposed to use trig for this

#

who would have thought

peak tide
#

yea that's a bit unorthodox but in principle there's no reason you can't do integration first

pulsar pagoda
#

my main issue with that book is not the topics, i would say its mathematical maturity

peak tide
#

i would characterize it as on the border between a calculus book and an analysis book

#

spivak is definitely more fun to read, but he doesn't do linear algebra or multivariable calculus or a bunch of the other special topics apostol does

pulsar pagoda
#

so your saying apostol is a better book as a reference i guess?

peak tide
#

it's perfectly ok to learn from, i personally didn't use it but i know for instance that the local university here uses it for their honors calculus sequence

pulsar pagoda
#

whats a honors course?

peak tide
#

basically the normal calculus sequence but a version aimed at math majors

pulsar pagoda
#

so intro to intro to real analysis?

peak tide
#

in particular, much more focus on proofs and doing things rigorously vs just learning to calculuate

peak tide
pulsar pagoda
#

sounds cool when u put it that way

#

xD

peak tide
#

it replaces the normal calculus sequence for those people who take it, but it's definitely good preparation for analysis

white spindle
#

@peak tide hey guys i got a thought on this, if we take a 42 neuron net, about the number of salient soul neurons in a bee or spider, than the typical human understanding of a flat neural network does not work because a 4x4 input is not dimensional enough for meaning. what we need for the bees vision in its 42 neurons is a 10x2 input in two sides from every antenna to a different neuron, with the remaining two neuron triggering the on off of the 10x2 sides so that the bee traces back and forth between the two sides of the two sides and then pyramidal neurons, of which the bee has 19, with some 54 neurotransmitters, control the response of the bee to have run specific spiking routines dependent upon chemical input induced also by its antennae. a bee has specific neuron assembly that allow high spike behavior to trigger a specific output - a bee's nervous system that is peripheral does most of the work that the cerebellum does in the human of translating valid spiking to correct motor - what this boils down to is that a bee is dependent on a specific soul signal and this can be somewhat generalized to humans!

peak tide
#

um

pulsar pagoda
#

hes trolling

peak tide
#

why did i get pinged for that haha

deep mango
white spindle
#

huh u said something about CNN

#

that is bee progs info i am working on

jagged forge
white spindle
#

huh all i was pointing out is that CNN isnt everything u got receptors and custom pyramidal behavior in biology

#

i got that in my lenet i am working on but it would cost lot of money to run it besides just as individual libraries - i am pretty poor

peak tide
#

sounds like a case for kickstarter

pulsar pagoda
#

man the hallucinations of chat gpt this days

#

their wild

peak tide
jagged forge
#

i honestly think he is not trolling

jagged forge
solid snow
#

52 factorial is scary

peak tide
#

you never swim in the same river twice, you never shuffle a deck of cards the same way twice

#

well unless you can do like 10^67 trials

#

it's a fun challenge in big data to even work out how you would detect that you shuffled it the same way twice

jagged forge
#

might as well wait for the universe to return to the same state as before you shuffled the cards

#

by the poincare recurrence theorem

peak tide
#

nice

jagged forge
#

actually nvm that’s a lot longer

#

apparently

peak tide
#

step 1: wait for universe to return to initial state

solid snow
#

i saw this golden reply to something about poincare years ago and its never left my mind

jagged forge
#

well, if it took less than the amount of times it takes to shuffle a deck 10^67 times

#

the people who care about shuffling a deck 10^67 times (which is probably an empty set) would care

white spindle
#

so, that is like a work of art in synchronizing what is there written down with what is there in the game, a lot like procedural generation has been given like an orbital or a molecule

#

i had alikened world of warcraft's orbital to water and starfields orbital to oxygen

solid snow
white spindle
#

that is all the orbital i got man, i am pretty sure the compound label is labeled right tho

peak tide
#

i'm not sure which is worse, the wall of incoherent text or that spastic animation that's gonna give me a seizure 😆

white spindle
#

huh, that is some text on what i have been doing, exploring the orbitals of specific video games which are hiding orbits

pulsar pagoda
#

the malware in that link

white spindle
#

huh? there is not malware there, just some hidden directories and shortcuts which might be what you are talking about

fringe summit
fringe summit
white spindle
#

well, it comes from ableton, and i have antimalware, so i doubt there is malware

pulsar pagoda
white spindle
#

not really just written in max, i know techno, he did his best but it was made in max

#

dont want to get all conspiratorial but definitely when you are using steves technology like PC/mac and iphone there is a malware risk, THAT IS NOT MY RISK

#

yea like i am going to definitely blame the modem and the hardware above anything that the rasmujin people did for steve to like intrinsically root windows or whatever

#

that is discosed, in leaks, dont want to point fingers but there was all kinds of government induced rooting on windows which presumably cleared up, who knows tho

sonic field
#

i'm confused

loud snow
#

what malware risk?

#

@white spindle government induced rooting on windows

#

this was win7

#

about government installing backdoors for win7 devices

neon elk
#

I do not want to study today. Can I take day off?

neat lintel
#

yes

neon elk
#

Ok. Ty.

flint basin
neat lintel
#

Are polar coordinates hard?

jagged forge
#

@pulsar pagoda sometimes databases become busy and can’t fulfill requests because of some background process they’re doing, eg indexing, compaction, snapshotting

#

sometimes they’re misconfigured which will cause it to do some of these things poorly or at an inopportune time

#

figuring out that that’s happening can be very difficult

pulsar pagoda
#

thats where automátion and monitoring becomes key

jagged forge
#

by busy i don’t necessarily mean it just shuts off, but you start to see poor performance

pulsar pagoda
#

yes

#

but it's hard to noticd

jagged forge
#

monitoring is important but if you don’t understand what the database’s metrics mean it doesn’t help opencry

pulsar pagoda
#

because u don't have a automated process which analysis the entire runtime

pulsar pagoda
#

disk usage and ram usage only gets u so far

#

u also need to monitor specific files

#

like socket to check network and such

#

i remember one time a index couldnt be loaded into ram

#

chaos

jagged forge
#

are you talking about like, those Unix “files” that show eg open file descriptors or connections

pulsar pagoda
#

yes

#

those ones

jagged forge
#

at my workplace we use prometheus, and there’s a premade Prometheus dashboard that exports metrics from the database. but when stuff goes wrong then we ssh into the node and start looking at stuff from the command line

pulsar pagoda
#

i actually don't dislike the infrastructure part of SE

jagged forge
#

it’s kind of like peering under the hood, isn’t it

pulsar pagoda
#

i wouldnt call it that

jagged forge
#

well, it’s not like hardware i guess lol

pulsar pagoda
#

but to me is where formalism matters because You can design architectures

#

software wise

#

and test if ur design resists the real world scenarios

#

sort of like software meets empiricism

jagged forge
#

yeah, they have things like TLA+ right

#

which has been used for distributed systems

#

among other things probably

pulsar pagoda
#

defining SLA

#

like

#

i know the idea

#

i just havent designed one

#

i would probably get to it

#

in My ingeniería de software course

jagged forge
#

SLA? isn’t that just like, uptime

pulsar pagoda
#

which i'm like taking

#

well SLA's are really important for SRE positions

#

let me Google tla that u mentioned

jagged forge
#

yeah SLA’s are important, i just wasn’t sure what you meant by designing one

pulsar pagoda
#

ohh

#

like designing systems to meet SLA's

#

tla is sort of like UML

#

and Z lang

#

yeah

#

those are cool

#

they teach UML at My uni

#

but i'm familiar from before

jagged forge
#

im not sure if UML is the right comparison

#

i think you write your algorithm in TLA and it finds bugs

pulsar pagoda
#

i'll read more

jagged forge
#

and it is particularly is used for concurrent or distributed systems

pulsar pagoda
#

wow

#

tla uses formal verification

#

jeez

jagged forge
#

yeah, hardcore stuff

pulsar pagoda
#

this goes beyond a language for design

#

this also helps u in designing and implementing test scenarios

jagged forge
#

invented by the same guy who invented LaTeX btw

#

leslie lamport

#

also known for lamport clocks

pulsar pagoda
#

who is a math guy

jagged forge
pulsar pagoda
#

why is it that math guys go to CS

#

and do cool shit

#

because CS is applied math

#

at least thats My take

#

CS is a branch of applied math

#

a really important one

jagged forge
#

yeah i mean software engineers are not computer scientists i guess

pulsar pagoda
#

yes

jagged forge
#

idk if there is an analogous example in another field

pulsar pagoda
#

computer engineers vs electronic engineers vs electrical engineers

jagged forge
#

engineers and physicists maybe? but not theoretical physicists

pulsar pagoda
#

electrical engineers and physicist

#

electrical engineering is similar to applied physics

jagged forge
#

interesting

pulsar pagoda
#

and things like CERN collider would be imposible with no engineers

#

so the guy i work with

#

works with physicist and math people in europe part time

#

he's the technical lead

#

so he tells them what they need tech wise

#

based on specs given by scientist

#

thats part of the schoolarship deal he got for doing his phd there

#

not in CERN though

pulsar pagoda
jagged forge
#

almost like you’re implying it’s closer to applied physics than regular engineering? idk

white spindle
#

yo electrical engineering is more military than anything else, you got mainly either the applied physics crowd at the foundry or more dedicated individuals esoterically aka anonymous

jagged forge
#

anonymous?

white spindle
#

you are really looking at manuals effectively when you are making a circuit work... i can mainly speak for silicon foundries in shenzao and you have tons of undescribed features in the circuit which makes EE a closed crowd like anonymous, even IIRC in page 2 of the intel CPU white paper it says very explicitly intel puts portions of the circuit down you do not have access to

pulsar pagoda
#

but it depends on the application

white spindle
#

i mainly engineer electric out of modular synthesizers because it is a challenge to use modular synthesizers in a whole new direction, for example i have 1 bit hash in modular, some basic RAM in modular, plus have experimented with affinities in modular where a specific small form factor effect (adv) is used as a control or detect to for a large form factor effect (adg)

pulsar pagoda
#

sort of what happens to mechanical engineers also

jagged forge
white spindle
#

huh, i had seen that you got some download cpu and access to silicon foundry that anonymous has which what my op is was related

#

like uh, to get some basic code done you basically need some custom silicon nowadays so that what this is basically modular

#

what is lurking behind live is what is the silicon foundry's secret which is the modular which makes everything work even irl

pulsar pagoda
#

@lament peak are u a mod?

white spindle
#

huh, i am kind of anon, i got u earlier access to something that the foundry might use

pulsar pagoda
#

maybe?

white spindle
#

huh like uh, modern, it is all frequency macross adders

pulsar pagoda
#

i mainly ignore this guy

#

nothing he says make sense to me

white spindle
#

naw uh, you got 10 frequencies the circuit hides, 1 frequency is exe and is hidden, 8 frequencies are the macross of the circuit and everything is encoded with delayed integral (direct)

#

for example we delay 333 666 999 and a monophonic impulse of 333 can be decoded

#

as well we may form a double full adder and have 1-8 behavior at frequency and this is a neuron

#

plus there is aforementioned the frequency macross adder which is the main power of the circuit which does direct training to the network of a specific nature

#

variously there are also high level process buffers which execute a process in series and chassis of high level process buffers which are what you see at the high level of the circuit on microscope

fresh stag
#

I hate graduate applications so much

#

So stressful man

vast surge
#

But yeah, it sucks

fresh stag
#

that channel is too professional for venting

#

I will restrict my whining to non stuck up strangers on the internet

#

as opposed to stuck up strangers on the internet

#

I see gradcafe posts with applications better than mine getting into like 1 school

#

or none

teal tree
weak vector
#

x=(3-y)/(y-1¹)

robust elbow
weak vector
#

what the

iron copper
rocky shuttle
#

ryc's an impersonator

robust elbow
flint basin
#

The real ryc is here

hasty peak
ocean harbor
#

not really his alt but probably a fun of ryc I can imagine that

little vine
#

I wonder what ryc thinks of his cult status

long matrix
#

negatively.

sonic sentinel
#

whwnever me or my friend make a joke that is moderately to severely less funny than expected we immediately respond with bullying

#

something along the lines of "not funny don't say this shit again"

#

i don't see it anywhere else i think that;s funny

noble osprey
#

hey maneee, sayyyy maneee, it's yaa boy Gangster Mathematician here🤙

pulsar pagoda
#

@jagged forge have u used oracledb?

#

honestly im amazed at its feature offering

#

plus docs

#

the cost keeps me from getting into it

#

but idk that it pays well to be a oracle DBA

jagged forge
pulsar pagoda
#

its a sql database

#

u know

#

oracle sql

jagged forge
#

what makes it better than free offerings

pulsar pagoda
#

but they been integrating other models

#

so they have integration of relational, non relational

#

and even blockchain based

#

dbs

#

in their main offering

#

offering a consistent api that allows for interoperability between all their diferent data models

jagged forge
#

interesting

pulsar pagoda
#

so if u had something like lets say an accounting ledger

#

u could implement it in ur distributed blockchain propietary db

#

and ensure desired properties

jagged forge
#

wait, you're saying you can hook your own db's into this thing?

pulsar pagoda
#

like double entry accounting

pulsar pagoda
jagged forge
#

i am not really familiar with multi-model databases

#

at my company we use many different databases for different things, but it also adds considerable overhead

#

but we don't pay for oracle database so

#

tradeoffs i guess

pulsar pagoda
#

yeah

#

still its a really interesting case of a niche thats being exploited in a specific market

#

sort of like what apples does with phones and laptops

#

really polished product

#

that is priced higher than market offerings