reinforcement-learning

[graceful widget]

Hello RL folks.
Seems like the space is not so active.

Anybody learning RL with Me

[silent acorn]

Hi I am new, can you learn with me?

[graceful widget]

Sure .
I'd be glad to

[graceful widget]

Are you reading any book for RL or taking courses

[silent acorn]

I learn from the course.

And you?

[graceful widget]

Reading a book for it.
I feel like the courses do not cover enough theoretical ground

[silent acorn]

Yes they teach me only Q-learning.
Can you say book name?

[icy tree]

Sutton and Barto, Introduction to Reinforcement Learning is a classic text, and it's also very good. Both for mathematical and theoretical treatment of the concepts.

[silent acorn]

Can you say where I get those books?

[icy tree]

https://inst.eecs.berkeley.edu/~cs188/sp20/assets/files/SuttonBartoIPRLBook2ndEd.pdf
from UC Berkeley

[silent acorn]

Okay thanks

[icy tree]

No problem!

[graceful widget]

Oh wow that's the same book I'm reading. It's a great text

[icy tree]

Though if you want a video course, David Silver's famous lectures are pretty good, nice balance of math

[silent acorn]

Can you give me link?

[icy tree]

Sure! This is the playlist by Google Deepmind YT channel:
https://youtube.com/playlist?list=PLqYmG7hTraZDM-OYHWgPebj2MfCFzFObQ&feature=shared

[sweet garden]

Hi,,iam Sharath ,,iam a beginner here in this field of data science, AI and machine learning,,I want to learn RL,,can anybody have time for group studies or guiding me??

[wise magnet]

👋

[graceful widget]

Yes. I do have some time for group study

@wise magnet nice to have you

[sweet garden]

Thanks 👍 looking forward to learn RL

[brazen mist]

Yes we can discuss here about deep reinforcement learning

As i finished a few topics and feeling confident but have some doubts as well

Any working professionals here ?

[graceful widget]

I'm not a working professional per se.
But I've got a good experience

[brazen mist]

Nice 🙂

[icy tree]

What doubts do you have?

[brazen mist]

There are a bunch of algorithms but don't know where to start to implement it

Also i want to implement those algo in games such as gta 5

[icy tree]

https://github.com/udacity/deep-reinforcement-learning
This might be useful
Good luck!

[feral sequoia]

Hi,
Did anyone work on multi agent algorithm in python ?

[fallow vigil]

do i ask rl related doubts here ?

[wet girder]

How can I balance the actor and critic networks?

[verbal latch]

Bhai pehle basics sikhle, uske bad gta and all ka bol

[brazen mist]

Bhai kay bolt reinforcement learning cha king hai me

[verbal latch]

I assume it's sarcastically said

[brazen mist]

No, I'm working professional

I have a speciality in model based deep reinforcement learning

[verbal latch]

Even Sutton and Barto or Lattimore won't claim the statement that you did

[brazen mist]

Yah i didn't say I'm superior, but I know how to solve problems using them , but yah still learning

[verbal latch]

Man it's not just about blindly using the algo, I see that you talk about Deep RL and stuff, and aim to do something crazy which is absolutely great, however the only concern is the basics.

Before even starting with Deep RL, there's a hell lot of stuffs involved in the learning and the planning setting, whether it be standard Bandits in pure exploratory/regret setting or MDP with various versions of it distributed across different types.

[marble sigil]

Hey there! I am a complete beginner to RL, I completed a beginner course on datacamp and solved the Frozen-Lake environment using what I learnt from that course, now I am trying to delve into Deep Q Learning and would like to know what resources I could potentially use, thanks!

[fiery jolt]

Why RL is so hard to understand and learn? Am I the only one having this issue?

[fiery jolt]

If so, can anyone tell me how to start with a simple explanation video or book or any other suggestions

[solar geyser]

https://www.youtube.com/watch?v=nyjbcRQ-uQ8&list=PLZbbT5o_s2xoWNVdDudn51XM8lOuZ_Njv

This is one of the best series I have seen on RL, she explains it really nicely

[fiery jolt]

Thank you

[solar geyser]

Yeah no problem man, hope this solves your problem

[fiery jolt]

Thanks for not sending david silver content or sutton book as I see the RL complicated even after seeing david silver lecture 2 and even with the book lol

But that is a video that I have never seen before so I know it is going to solve my problems for sure

[solar geyser]

Yeah sending an entire book seems like overkill especially if you just want a foundation

[solar geyser]

😂

[fiery jolt]

Yeah, the book will overwhelm you with a ton of theories and equations that you don’t need but will just make you suffer and depressed on the RL

[solar geyser]

Real

[fiery jolt]

Thanks man for the amazing playlist you send above I finished 7 of the 15 videos on the playlist I feel like I know more than I could possibly imagine in reinforcement learning

And I will definitely watch the rest to make projects in RL and Deep RL

[solar geyser]

That’s great to hear bro, glad you learned a lot

😁

[fiery jolt]

It’s all thanks to you bro❤️

[solar geyser]

Lmk if you need anything else man

[fiery jolt]

Let’s do some RL projects sometime and learn from each other

[verbal latch]

Those interested in bandit literature can check our recent work out

http://arxiv.org/abs/2408.14195

TLDR: We analyse a clustered multi-armed bandit formulation, where the learning objective is to identify representative arms from each cluster, in a fixed confidence setting

@here

[silent siren]

Huh can i see it too pls ?

[fiery jolt]

https://www.youtube.com/watch?v=nyjbcRQ-uQ8&list=PLZbbT5o_s2xoWNVdDudn51XM8lOuZ_Njv

Here is the playlist

[silent siren]

https://youtu.be/E0j680JoBko

[teal folio]

As compared to conventional ML and DL it is because it requires software development sort of coding techniques to be used. While ML and DL has simple libraries that you can just simply import and work with

[silent siren]

https://youtu.be/7NWRnWZghGA

[silent siren]

https://youtu.be/pPeiwUAVl08?si=eidOoW7JFa_NknaE

[loud harbor]

ruhaan10 has been warned

Reason: Bad word usage

[dry widget]

Does anyone have best reinforcement learning playlist with maths

[tired cove]

@dry widget Check out UCLxDeepmind RL series on youtube. You can also find a specialization from Uni of alberta on coursera. You can audit for free

[dry widget]

Thanks 🌹

[native kernel]

i dont know too much about reinforcement learning, but is it possible to make a program that takes cards from a trading card game (like pokemon or MTG) and figure out what a very good deck is by having it battle other decks?

i ask if it's possible because i dont know if it would take too long to run and if it would need like millions of battles to figure it out

[normal siren]

Yes, you can have a look at https://rlcard.org/ as an example

[normal siren]

Depending of the game, here they report it could take ~500k steps to train the agent: https://doi.org/10.48550/arXiv.1910.04376

[native kernel]

@normal siren oh wow thank you!! I'll check that out

[tender seal]

HI, I am Abdullah I am an ML engineer want to join any team to particapte in kaggle competions

[spice wren]

i need help to take this test. can we google meet please to help me out?

[loud palm]

Hi, Please help me.
I'm going to make a search engine based on customer behaviors.
Inputs: query embedding, history embedding (Metadata is stored with vector formats)
We use cosine similarity and train embedding models using multi armed bandits.(Is that possible?)
I have two questions about this.
First, how to get the gradient of embedding when use consime similarity?(Can that be estimated in torch?)
Second, for the search, we use two steps, updating the weights about historical embedding and query embedding at the same time, I think that can be noisy.
But I can't make sure. I attahced diagram. And if any questions, feel free to ask.
https://drive.google.com/file/d/1_vWxdasnHjCL6_momviQzcDYHAAgc-1M/view?usp=sharing

[verbal otter]

I believe this channel is about Discussing Reinforcement learning kindly refrain to put any other content or stuff which is unrelated

[merry pollen]

Hello everyone, kindly suggest me best courses on reinforcement learning and share reviews about reinforcement learning specialization on Coursera

[normal siren]

https://bsky.app/profile/sungkim.bsky.social/post/3lcvvmtmcoc2u

Sung Kim (@sungkim.bsky.social)

Some learning resources on Reinforcement Learning in no particular order.

• UCL Course on RL ( www.davidsilver.uk/teaching/ )
• UC Berkeley's Deep Reinforcement Learning ( rail.eecs.berkeley.edu/deeprlcourse/ )
• Lil's Reinforcement-Learning ( lilianweng.github.io/tags/reinfor... )

Bluesky · Dec 9, 2024, 11:41 PM

[merry pollen]

Thanks

[normal siren]

Some more material:

• http://incompleteideas.net/book/the-book-2nd.html
• https://huggingface.co/learn/deep-rl-course/unit0/introduction
• https://doi.org/10.48550/arXiv.2412.05265
• https://www.davidsilver.uk/teaching/
• https://gibberblot.github.io/rl-notes/index.html

[merry pollen]

Thank you 💕

[normal siren]

Forgot this one: https://spinningup.openai.com/

[spice cypress]

anyone interested in teaming up for lux AI ?

[gloomy token]

How good is DQN method

For EHR'S

[fallen cargo]

Anyone used NotebookLM how is it?

[fallen cargo]

Hi I am new here I think my earlier message was not relevant to this thread please tell me in which thread I can ask such questions
Thank you

[jagged oriole]

https://discord.com/channels/1101210829807956100/1129507816697241822

[clear mortar]

Hello, I am learning Reinforcement Learning and interested in Automation, robotics and automotive technologies. Looking for a peer or group to learn together. Is anyone interested?

[verbal otter]

Hi Everyone, I am a 2nd Year PhD student in Computer Science at University of Maryland Baltimore County specializing in Machine Learning, Reinforcement Learning, and Mathematical Reasoning in LLMs. I was thinking to write a Review paper on the current Maths Reasoning in LLMs , so was looking for potential collabrators on it. Thanks

[spice cypress]

YES

[ember blaze]

Hi, I learn supervised learning and unsupervised learning, I Interesting to talk together

[fiery jolt]

Yes I am very interested

[wise canopy]

Hi I am also learning automation and robotics. I am interested to join!

[past cloud]

yes I am interested to join

[white olive]

This Python class offers a multiprocessing-powered Pool for efficiently collecting and managing experience replay data in reinforcement learning.

https://github.com/NoteDance/Pool

[boreal dagger]

Hi. I'm starting with RL, namely PPO and by extension GRPO. Anyone has prior experience?!

[final sphinx]

im working on a project involving RL and drone delivery optimization, could you guys help by sharing some resources for learning RL that actually helped you all in learning RL? Thanks!

[sinful turtle]

Job Title: Part-Time Senior AI/ML Engineer (Remote)

We are seeking a skilled and experienced Senior AI/ML Engineer to join our remote team on a part-time basis. The ideal candidate will have a strong technical background, excellent communication skills, and the ability to work independently in a fast-paced environment.

Requirements:
-Minimum of 7–10 years of professional software development experience

-Proven experience working effectively in a remote environment

-Advanced English proficiency (C1 or higher); an American accent is preferred

-Availability to work 10–15 hours per week during EST or CST business hours

If you're a highly motivated engineer with a passion for building high-quality software and can commit to a flexible part-time schedule, we’d love to hear from you.
You can connect with me on WhatsApp: [phone]

[bold arch]

Hi, @everybody
I have one question, I'm training ml models for the prediction, which is classification problem of 3 classes, where the number of samples are similar but the predition is skewed.
First class and second class is predicted with low precision tough, third class is never predicted. What's the reason? I can' t find the reason.
Before, when I applyed reinforcement learning, where the three classes were assigned to three actions and one action is never selected, too.
Actually, that is the preeiction model of forex eur/usd.

[tidal notch]

https://media.discordapp.net/attachments/1436719817624256534/1436719913518633010/1.JPG?ex=6910a130&is=690f4fb0&hm=6a48397700e40b701b7defba0bc73ccc590e83e58af09eb7035cae318e9fb319&=&format=webp&width=515&height=687
https://media.discordapp.net/attachments/1436719817624256534/1436719914034659408/2.jpg?ex=6910a130&is=690f4fb0&hm=5d3c01e3db0b2fe7135969c69c22cbf49db07bae5ed8cb9a98ac3e18d3c73ce5&=&format=webp&width=515&height=687
https://media.discordapp.net/attachments/1436719817624256534/1436719914512547951/3.jpg?ex=6910a130&is=690f4fb0&hm=59a326eaa4d74733a406431b5c2eb8ee07f6b78d95094102deb1153d2e261407&=&format=webp&width=515&height=687

[weak jewel]

me

[tawdry yarrow]

The reason might be a Class imbalance in the Training Data, make your Training data class balanced , that all three classes have the same amount of samples in the whole of the training data.

[bold arch]

I'm finding a US developer for the collaboration. If anybody interested, please dm me.

[gleaming gorge]

https://www.kaggle.com/code/petrumihaicraciun/buck-shot-roulette-reinforcement-learning

Quick little game thing

if anyone has played Buck Shot roullete this is a little sample enviroment I made that you can make a AI on

very similar to the video game

[solar knoll]

@gleaming gorge This is really cool, love that you used Mesa for the multi-agent setup, makes it super easy to swap in different agent classes. The BaseAgent abstraction is clean too.

[gleaming gorge]

Now that I'm playing around with my PhD more I also recommend looking at @dataclasses in python for large models

I've kind of started moving it to seperate files because after 500 lines of code mesa can get hard to track so moving it into a kind of variable holder can help

https://docs.python.org/3/library/dataclasses.html