sqIRL Lab

The deadline for the #AIMLAI workshop held jointly with #ECMLPKDD2025 has been extended until June 21st.
Looking forward to last-minute submissions on work around #interpretability and #explainability of #AI / #ML

project.inria.fr/aimla
#mechinterp #xai

Part of the #sqIRL lab at the IDLab day 2025 #uantwerp

Our lab got two papers accepted at #ECMLPKDD2025 on the topics of #Interpretability for Spiking NNs and self-supervised representation learning with embedded interpretability .
Congrats to Jasper, Hamed, Fabian and our collaborators.

#SNN #SIM #AI #ML #neuromorphic #xai #interpretableML

Honored to be selected among the Outstanding Reviewers at #CVPR2025.
#UAntwerp @sqirllab.bsky.social #IDLab

cvpr.thecvf.com/Conferences/...

It is confirmed, the #AIMLAI workshop will be held jointly with @ecmlpkdd.org.
We invite the submissions of long and short papers covering work around #interpretability and #explainability of #AI/#ML.

Deadline: 14/06/25
CfP: shorturl.at/yYQ9G
Website: shorturl.at/W9r1A

#XAI #mechinterp #ECMLPKDD

Last week our lab celebrated the doctoral defense of Hamed Behzadi. It has been four years since Hamed join us, and his evolution into fully fledged independent researcher has been constant. Congratulations!

See shorturl.at/296MV for some of the work produced by Hamed.

#ML #AI #Interpretability

Improving Neural Network Accuracy by Concurrently Training with a... Recently within Spiking Neural Networks, a method called Twin Network Augmentation (TNA) has been introduced. This technique claims to improve the validation accuracy of a Spiking Neural Network...

Benjamin Vandersmissen (25/04 evening) will show the effects that using a twin network has on learning processes and share insights on how TNA leads to superior predictive performance in a number of tasks for several architectures. #deeplearning #ML #ICLR2025 #sqIRL
openreview.net/forum?id=TEm...

Bilinear MLPs enable weight-based mechanistic interpretability A mechanistic understanding of how MLPs do computation in deep neural net- works remains elusive. Current interpretability work can extract features from hidden activations over an input dataset...

Thomas Dooms will show how to bilinear MLPs can server as more transparent component that provides a better lens to study the relationships between inputs, outputs and the weights that define the models. #mechinterp #interpretability #ML #AI #XAI #ICLR2025 #sqIRL
openreview.net/forum?id=gI0...

If you are at #ICLR2025 and interested on how to understand DNNs from its weights and on how to improve predictive performance of a DNN via Twin Network Augmentation, we encourage you to get in touch with Thomas and Benjamin who will be presenting our work there. #sqIRL #UAntwerp #XAI

We had the opportunity to contribute to the Research Day of the Antwerp Center of Responsible #AI ( #ACRAI ) where Salma and Hamed presented their work on #explainability-driven #HSI analysis and model #interpretability, respectively.
#ML @uantwerpen.be
www.uantwerpen.be/en/research-...

This week we had the visit of Prof. Eliana Pastor (DBDMG @PoliTO) who gave a presentation on her research around the topics of #trustworthyAI, #Bias analysis and #FairnessAI. Very good work and interesting ideas. @elianapastor.bsky.social we hope to host you again soon. #explainability #AI #ML

Bilinear MLPs enable weight-based mechanistic interpretability A mechanistic understanding of how MLPs do computation in deep neural networks remains elusive. Current interpretability work can extract features from hidden activations over an input dataset but gen...

Bilinear MLPs Enable Weight-based Mechanistic Interpretability
M. Pearce, T. Dooms, A. Rigg, J. Oramas, L. Sharkey

We show that bilinear layers can serve as an interpretable replacement for current activation functions; enabling weight-based interpretability.

preprint: arxiv.org/abs/2410.08417

Improving Neural Network Accuracy by Concurrently Training with a... Recently within Spiking Neural Networks, a method called Twin Network Augmentation (TNA) has been introduced. This technique claims to improve the validation accuracy of a Spiking Neural Network...

Improving Neural Network Accuracy by Concurrently Training with a Twin Network
B. Vandersmissen, L. Deckers, J. Oramas

We show the effectiveness of TNA lies on a better exploration of the parameter space and the learning of more robust and diverse features

preprint: openreview.net/forum?id=TEm...

A great start for 2025.
Proud to announce that our group (#sqIRL/IDLab) got two papers accepted at #ICLR2025. A first for our young lab.

Thanks to our collaborators, the FAIR Program and the Dept. of CS @uantwerpen.bsky.social for supporting this research.

#AI #ML #interpretability #XAI

Recent work published by the #sqIRL Lab on the training of competitive deeper Forward-Forward Networks. #FF #localLearning #ML #RepresentationLearning

AI Conferences @neuripsconf.bsky.social @cvprconference.bsky.social @iccv.bsky.social @eccv.bsky.social @iclr-conf.bsky.social @blog.neurips.cc.web.brid.gy

go.bsky.app/45EuhSi

Neural networks are not black boxes. They're the opposite of black boxes: we have extensive access to their internals.

I think people have accepted this framing so innately that they've forgotten it's not true and it even warps how they do experiments.

A starter pack of people working on interpretability / explainability of all kinds, using theoretical and/or empirical approaches.

Reply or DM if you want to be added, and help me reach others!

go.bsky.app/DZv6TSS

50-day Free Access Link: eur01.safelinks.protection.outlook.com?url=https%3A...

Towards the characterization of representations learned via capsule-based network architectures Capsule Neural Networks (CapsNets) have been re-introduced as a more compact and interpretable alternative to standard deep neural networks. While rec…

Congratulations to Saja Tawalbeh on the acceptance of her work on the #interpretability of #Capsule Networks at the #Neurocomputing Journal.
Pre-proof: www.sciencedirect.com/science/arti...
@uantwerpen.bsky.social #XAI #ML #AI #sqIRL

Check out this starter pack!

go.bsky.app/BYkRryU