Lightnews — Scholar-powered news

Reposted by Kale-ab Tessera

RL & Agents Reading Group @rl-agents-rg.bsky.social · Sep 3

📢 RL reading group Thursday @ 16:00 BST 📢

Speaker: Alex Lewandowski

Title: The World Is Bigger: A Computationally-Embedded Perspective on the Big World Hypothesis 🌍

Details: edinburgh-rl.github.io/reading-group

UoE RL Reading Group

University of Edinburgh Reinforcement Learning Reading Group

edinburgh-rl.github.io

3 6

Kale-ab Tessera @kale-ab.bsky.social · Aug 19

Refreshing to see posts like this compared to "we have 15 papers accepted at X" 🙌

1 1

Reposted by Kale-ab Tessera

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · Aug 18

None of our impactful papers have had an easy path through traditional venues.
Most cited paper? Rejected four times.
Most impactful paper? Poster at a conference.
But none of it matters because arxiv makes everything work

6 6 110

Kale-ab Tessera @kale-ab.bsky.social · Aug 18

Great first couple of days at DLI @deeplearningindaba.bsky.social in Kigali 🇷🇼, some highlights include amazing talks talks by @verenarieser.bsky.social and Max Welling, great pracs and tuts, and of course the opening party ( before the rain 😢) 🎉 #DLI2025

1 4

Reposted by Kale-ab Tessera

Deep Learning Indaba @deeplearningindaba.bsky.social · Aug 17

We’re excited to unveil the first #DLI2025 lineup of tutorials and practicals:

✨ Machine Learning Foundations
✨ Generative Models & LLMs for African languages

All tutorial content will also be available online after the Indaba. Don’t miss out, subscribe here 👉 lnkd.in/eCgXRqsV

3 2

Kale-ab Tessera @kale-ab.bsky.social · Aug 3

🙌🎉

Kale-ab Tessera @kale-ab.bsky.social · Aug 3

🇨🇦 Heading to @rl-conference.bsky.social next week to present HyperMARL (@cocomarl-workshop.bsky.social) and Remember Markov (Finding The Frame Workshop).

If you are around, hmu, happy to chat about Multi-Agent Systems (MARL, agentic systems), open-endedness, environments, or anything related! 🎉

2 9

Reposted by Kale-ab Tessera

Deep Learning Indaba @deeplearningindaba.bsky.social · Jul 30

We are thrilled to announce our next keynote speaker
@wellingmax.bsky.social, Professor at the University of Amsterdam, Visiting Professor at Caltech and CTO & Co-Founder of CuspAI.
Catch his talk “How AI could transform the sciences” on August 18 at 4:30 PM GMT+2.
#DLI2025

1 1

Reposted by Kale-ab Tessera

RL & Agents Reading Group @rl-agents-rg.bsky.social · Jul 24

RL reading group TODAY @ 15:00 BST 🔥

Speaker: Cam Allen (Postdoc, UC Berkeley)

Title: The Agent Must Choose the Problem Model

Details: edinburgh-rl.github.io/reading-group

UoE RL Reading Group

University of Edinburgh Reinforcement Learning Reading Group

edinburgh-rl.github.io

1 3

Kale-ab Tessera @kale-ab.bsky.social · Jul 23

Always nice to see when simpler methods + good evaluations > more complicated ones. 👌

1

Kale-ab Tessera @kale-ab.bsky.social · Jul 10

Reading group is back for those interested in RL/MARL/agents/open-endedness and alike... First session today at 3pm BST, @mattieml.bsky.social is presenting the Simplifying TD learning/PQN paper. 🎉 Meeting link: bit.ly/4lfdaGR Sign up: bit.ly/40xNQDR

RL & Agents Reading Group @rl-agents-rg.bsky.social · Jul 10

We are super excited to kick things off again with Mattie Fellows (Postdoc @ FLAIR in Oxford) today 15:00 BST!

Paper: Simplifying Deep Temporal Difference Learning

Check out our website for full info edinburgh-rl.github.io/reading-grou...

1 3

Reposted by Kale-ab Tessera

RL & Agents Reading Group @rl-agents-rg.bsky.social · Jul 10

Hello world! This is the RL & Agents Reading Group

We organise regular meetings to discuss recent papers in Reinforcement Learning (RL), Multi-Agent RL and related areas (open-ended learning, LLM agents, robotics, etc).

Meetings take place online and are open to everyone 😊

1 12 37

Kale-ab Tessera @kale-ab.bsky.social · Jun 24

This has happened to me too many times 🤦‍♂️ Also doesn't help that Jax and PyTorch use different default initialisations for dense layers.

2

Kale-ab Tessera @kale-ab.bsky.social · Jun 23

Well done & well deserved!! 🎉🎉 It has been awesome to see this project evolve from the early days.

1 1

Kale-ab Tessera @kale-ab.bsky.social · Jun 5

The Edinburgh one will be back and running soon. We are just updating the website and other things. There is this form for people interested - forms.gle/DAbkpN9b4cUt...

Edinburgh RL Reading Group

Please add your details so that you can remain on the mailing list for the RL Reading Group.

forms.gle

3

Kale-ab Tessera @kale-ab.bsky.social · May 28

Forgot to also add ⚡ quickstart link for people who like to experiment on notebooks: github.com/KaleabTesser...

github.com

1

Kale-ab Tessera @kale-ab.bsky.social · May 28

Thanks for checking it out! 👍 Good point, there might be an interesting link between MoEs and hypernets. We used hypernets since they're simpler (no need to pick or combine experts), and maximally expressive (gen weights directly).

Lol yes, will had a .gitignore, missed it when copying things over.

1

Kale-ab Tessera @kale-ab.bsky.social · May 27

🎯 TL;DR: HyperMARL is a versatile approach for adaptive MARL -- no changes to the RL objective, preset diversity, or seq. updates needed. See paper & code below!

Work with Arrasy Rahman, Amos Storkey & Stefano Albrecht.

📜: arxiv.org/abs/2412.04233
👩‍💻: github.com/KaleabTessera/HyperMARL

HyperMARL: Adaptive Hypernetworks for Multi-Agent RL

Adaptability to specialised or homogeneous behaviours is critical in cooperative multi-agent reinforcement learning (MARL). Parameter sharing (PS) techniques, common for efficient adaptation, often li...

arxiv.org

1 3

Kale-ab Tessera @kale-ab.bsky.social · May 27

⚠️ Limitations (+opportunity): HyperMARL uses vanilla hypernets, which can inc. param. count esp. MLP hypernets. In RL/MARL this matters less (actor-critic nets are small), and params grow ~const with #agents, so scaling remains strong. Future work could explore chunked hypernets.

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

🔎 We also do ablations and see the importance of the decoupling and the simple initialisation scheme we follow.

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

📊 We validate HyperMARL across various diverse envs (18 settings; up to 20 agents) and find that it achieves competitive mean episode returns compared to NoPS, FuPS, and modern diversity-focused methods -- without using diversity losses, preset diversity levels or seq. updates.

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

💡To address the coupling problem, we propose 𝐇𝐲𝐩𝐞𝐫𝐌𝐀𝐑𝐋: a method that explicitly 𝐝𝐞𝐜𝐨𝐮𝐩𝐥𝐞𝐬 obs- and agent-conditioned gradients with hypernetworks. This means obs grad noise is avg. per agent (Zᵢ) before applying agent-cond. grads (Jᵢ) -- unlike FuPS, which entangles both.

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

🔬 We isolate FuPS’s failure in matrix games: shared policies struggle when agents need to act differently. Inter-agent gradient interference is at play -- especially when obs and agent IDs are 𝐜𝐨𝐮𝐩𝐥𝐞𝐝. Surprisingly, using only IDs (no obs) performed better and reduced interference.

Performance and gradient interference plots.

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

❓Existing methods add a diversity loss, use sequential updates or require knowing the optimal task diversity level beforehand. These can be hard to tune or inefficient. We ask: can shared policies adapt without any of the above?

1 1

Kale-ab Tessera @kale-ab.bsky.social · May 27

⚖️ 𝐖𝐡𝐚𝐭’𝐬 𝐭𝐡𝐞 𝐢𝐬𝐬𝐮𝐞? In MARL, optimal performance requires representing the right behaviours. Separate networks per agent (NoPS) enable agent specialisation but is costly & sample-inefficient; shared networks (FuPS) are efficient but lack agent diversity/specialisation.

1 1