Mathieu Chalvidal (@ChalviM) profile picture

Ph.D candidate at ANITI - Artificial and Natural Intelligence Toulouse Institute and Brown University

Toulouse - France
Joined October 2014
Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)Photo by Mathieu Chalvidal (@ChalviM)
Sure, sgd is good, but have you tried learning your own learning rule? In a new #NeurIPS2022 paper (arxiv.org/abs/2202.02363), @Tserre, R. Vanrullen and I explore whether meta-learnt synaptic dynamics could encode an efficient reinforcement learning program. Spoiler, they can...
2
11
1
81
Mathieu Chalvidal retweeted
this is so bad ...
5
29
3
84
19,101
Mathieu Chalvidal retweeted
Let's do TLDR for our new preprint! Brain-Diffuser arxiv.org/abs/2303.05334 We reconstructed perceived scene images from brain signals. These reconstructions capture both layout and semantic information of images... But how?
4
51
14
229
88,211
New optimization paper! 〽️🎢 Convergence guarantees and acceleration ratios for the Block Delayed Majorize-Minimize Memory Subspace Gradient descent algorithm, which solves centralized federated learning problems with asynchronous distributed updates! iopscience.iop.org/article/1…
1
149
Mathieu Chalvidal retweeted
GANs aren't very creative; VAE lack quality; but diffusion models strike a goldilocks zone, being both creative and generating high-quality sketches. Still a bit of a gap compared to humans, but it's closing in. Intriguing work from @VictorBoutin @tserre arxiv.org/abs/2301.11722
1
26
1
129
17,029
...We crafted networks able to continually self-modify themselves through recursive updates as a function of their weight state and sensorial information. The resulting agents are able to adapt fast and precisely in multiple tasks from maze navigation to robot control...
2
5
If you are interested to chat about this work and meta-learning in general, I'll be in NOLA this week. Poster session is Thu. at 2 p.m. PST. Looking forward to it!
...Interesting properties emerge: Agents adapt rapidly, with behavioural changes interpretable as sharp modulations of the network Hopfield energy and identifiable attractors in the synaptic space .They are more robust and learning transfer better to unseen tasks variations...
5
Mathieu Chalvidal retweeted
We are pleased to present our work: an efficient attribution method based on Sobol indices for explainability! Our new NeurIPS paper is available at arxiv.org/abs/2111.04138. github.com/fel-thomas/Sobol-… for the implementation of the method🎉
Preparing the next #NeurIPS2021 conf. Have a look at our 4 papers! (1/4) Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis by Thomas Fel, @RemiCadene, Mathieu Chalvidal, @quobbe, David Vigouroux, Thomas Serre arxiv.org/abs/2111.04138
1
4
8
Mathieu Chalvidal retweeted
To appear at NeurIPS 2020 (spotlight). Paper, code, and data at serre-lab.github.io/crbpsite…. Thanks to the reviewers for suggesting we emphasize the flood-filling routine discovered by the models for segmentation, resembling classic neuro/cog-sci work on human segmentation routines.
Stable and expressive recurrent vision models New work from the group with @alekhka, @lakshming92, @RexGLiu, and @tserre. project website: serre-lab.github.io/crbpsite… arxiv: arxiv.org/abs/2005.11362 Read on for an explainer!
2
16
1
37
Excited to present my first work as a PhD student at @ANITI_Toulouse and @tserre-lab at @BrownUniversity with Rufin VanRullen and Thomas Serre: "Neural Optimal Control for Representation Learning". Preprint arxiv.org/abs/2006.09545 Code & Notebook to come! Read more below! 1/9
1
20
2
60
Overall, this work suggest that a static parametrization of ANN is suboptimal as it constrains a single transformation to account for the whole data pop. On the contrary, considering them as flexible objects dynamically modulated might enhance their representation power. 8/9
1
2
In future work, we would like to investigate the robustness and generalization properties of such controlled models as well as their similarities with fast-synaptic modulation systems observed in neuroscience. 9/9 Thanks for reading!
1