Mirco Musolesi
@mircomusolesi.bsky.social
120 followers 150 following 48 posts
Professor of Computer Science. Machine Intelligence Lab, UCL AI Centre, Department of Computer Science, University College London. Artificial Intelligence | Machine Learning | Computing | Books https://www.mircomusolesi.org
Posts Media Videos Starter Packs
mircomusolesi.bsky.social
New preprint: “Complexity-Driven Policy Optimization”. The paper discusses a novel complexity-driven policy optimisation solution for efficient and robust approximation in Reinforcement Learning. Essentially, entropy alone is not sufficient, we also need "structure".
gfranceschelli.bsky.social
📢 New Paper!

We replace the entropy bonus in PPO with a *complexity* bonus, encouraging structured and stochastic policies that are robust to different scaling factors and can work in environments with variable exploration needs.

Read more:
arxiv.org/abs/2509.20509

w/ @mircomusolesi.bsky.social
Complexity-Driven Policy Optimization
Policy gradient methods often balance exploitation and exploration via entropy maximization. However, maximizing entropy pushes the policy towards a uniform random distribution, which represents an un...
arxiv.org
mircomusolesi.bsky.social
Received a copy of “The Library of Lost Maps” by my colleague James Cheshire at @uclgeography.bsky.social in the internal mail. Really beautiful (and very interesting) book. Thanks a lot James!
mircomusolesi.bsky.social
For the geeks among you: just discovered the Apple Calculator app has a "Programmer" mode, complete with binary, hex, and related operations.
mircomusolesi.bsky.social
The problem is still relevant - we now have Amazon delivery vans instead to play with. 2/2
mircomusolesi.bsky.social
I was thinking about this some weeks ago. The Travelling Salesman Problem (finding the shortest path through a list of cities, visiting each only once), a classic in Computer Science/Maths, is becoming something of the past without actual salesmen on the roads. 1/2
economist.com
In the 1980s and 1990s salesmen slogged up and down Britain’s motorways. Today that “road-warrior” lifestyle has disappeared. There are several reasons why
British men are driving less, and a culture is vanishing
Farewell to the road warriors
econ.st
mircomusolesi.bsky.social
Wonders of the natural world. I inherited an office cactus from a colleague who left. In a matter of days, out of nothing the cactus has developed a very long stem (with flowers).
Cactus.
mircomusolesi.bsky.social
Today Charlie Westphal will present our paper "Feature Selection for Network Intrusion Detection" in the main Research Track of ACM KDD 2025 in Toronto.

Link to the paper: dl.acm.org/doi/pdf/10.1...

#KDD2025
mircomusolesi.bsky.social
Liza Tennant (liza-tennant.github.io) and her #RLDM2025 poster on “Dynamics of Moral Behaviour in Heterogenous Populations of Learning Agents".

Paper: ojs.aaai.org/index.php/AI...
mircomusolesi.bsky.social
Lorenz Wolf (profiles.ucl.ac.uk/90553-lorenz...) and his #RLDM2025 poster on “Heterogeneous Knowledge for Augmented Modular Reinforcement Learning”.

Preprint: arxiv.org/abs/2306.01158
mircomusolesi.bsky.social
If you are attending #RLDM2025 in Dublin this week, do not miss the posters by Liza Tennant (liza-tennant.github.io) and Lorenz Wolf (profiles.ucl.ac.uk/90553-lorenz...) showcasing the work of our lab!

rldm.org

@rldmdublin2025.bsky.social
mircomusolesi.bsky.social
"Mr. Atkinson is credited with inventing many of the key aspects of graphical computing, such as “pull down” menus, and popularizing the “double-click” gesture, which allows users to open files, folders and applications by clicking a mouse button twice in succession..." 2/2
mircomusolesi.bsky.social
Many posts about Apple announcements following WWDC25. I definitely like the glassy widgets. Less attention has been paid to the passing of Bill Atkinson, a real (personal) computer pioneer. This NYT obituary is quite good: www.nytimes.com/2025/06/07/t... 1/2
Bill Atkinson, Who Made Computers Easier to Use, Is Dead at 74
www.nytimes.com
mircomusolesi.bsky.social
8 days left to apply!
mircomusolesi.bsky.social
The second call for applications to the CDT in Cyber-Physical Risk is now open. I am offering two projects in the area of AI and cybersecurity. I would be happy to chat about this opportunity if you are interested.

www.ucl.ac.uk/security-cri...

Closing date: 15 June 2025.
mircomusolesi.bsky.social
The second call for applications to the CDT in Cyber-Physical Risk is now open. I am offering two projects in the area of AI and cybersecurity. I would be happy to chat about this opportunity if you are interested.

www.ucl.ac.uk/security-cri...

Closing date: 15 June 2025.
mircomusolesi.bsky.social
Also, presented in the Blue Sky Track, so this seems the right place to advertise it.
mircomusolesi.bsky.social
Today our paper with James Rudd-Jones and María Pérez-Ortiz "Multi-Agent Reinforcement Learning Simulation for Environmental Policy Synthesis" will be presented at the #AAMAS2025 Conference in Detroit.

Time: 2pm ET
Room: Richard

Link to the paper: arxiv.org/abs/2504.12777

@aamasconf.bsky.social
mircomusolesi.bsky.social
Today our paper with James Rudd-Jones and María Pérez-Ortiz "Multi-Agent Reinforcement Learning Simulation for Environmental Policy Synthesis" will be presented at the #AAMAS2025 Conference in Detroit.

Time: 2pm ET
Room: Richard

Link to the paper: arxiv.org/abs/2504.12777

@aamasconf.bsky.social
mircomusolesi.bsky.social
Congratulations to Liza Tennant (liza-tennant.github.io) for passing her PhD viva today with a thesis on “Moral Alignment of Agentic AI Systems”! Well done Liza! And thanks to the examiners @jzleibo.bsky.social and John Shawe-Taylor for examining Liza’s dissertation.
mircomusolesi.bsky.social
If you are attending #AISTAT2025, do not miss Charlie's presentation tomorrow.
mircomusolesi.bsky.social
Our paper "Partial Information Decomposition for Data Interpretability and Feature Selection" with Charlie Westphal and Steve Hailes will be presented at #AISTAT2025 on Sunday.

Session details: Spot 151 in Session 2 (Sun, May 4).

Link to the paper: openreview.net/attachment?i...
mircomusolesi.bsky.social
Our paper "Partial Information Decomposition for Data Interpretability and Feature Selection" with Charlie Westphal and Steve Hailes will be presented at #AISTAT2025 on Sunday.

Session details: Spot 151 in Session 2 (Sun, May 4).

Link to the paper: openreview.net/attachment?i...
mircomusolesi.bsky.social
Our paper on model-based reinforcement learning for causal discovery is now out in Proceedings of the Royal Society A.

@royalsocietypublishing.org
@uclengineering.bsky.social
vadarvariu.bsky.social
Our paper "Tree search in DAG space with model-based reinforcement learning for causal discovery" has just been published in Proceedings of the Royal Society A. Joint work with Steve Hailes and @mircomusolesi.bsky.social 🧵 1/
Screenshot of the paper title, authors, abstract and miscellaneous bibliographical information as it appears in the published journal PDF.
mircomusolesi.bsky.social
Having been quite interested in the idea of the Animal-AI Olympics (github.com/Kinds-of-Int...) in the past, I am very happy to see that ICLR is colocated with PetExpo this year. #ICLR2025
mircomusolesi.bsky.social
This afternoon Liza Tennant will present our paper "Moral Alignment for LLM Agents" at @iclr-conf.bsky.social 2025.

Time: 3pm Singapore time
Location: Hall 3 + Hall 2B #238

Paper: openreview.net/pdf?id=MeGDm...

Poster+video: iclr.cc/virtual/2025...

#ICLR2025 #ICLR25
mircomusolesi.bsky.social
...it took me a while to find out that this was the cause of the problem.