"Leveraging Gradient Noise for Detection and Filtering of Byzantine Clients" by Kungurtsev and colleagues, exploring the use of higher order moment statistics for identifying Byzantine clients in distributed learning, has been accepted for publication in IEEE Access.

"Probabilistic Iterative Hard Thresholding for Sparse Learning", by Kungurtsev and co-authors, presenting a method for sparsity-promoting estimation with strong convergence guarantees, has been accepted for publication in Computational Optimization and Applications.

"Optimal Control of Two-Phase Membrane Problem" by Bozorgnia and Kungurtsev, studying the challenging obstacle problem exhibiting operator nonsmoothness, has been accepted for publication in the journal of Applied Mathematics and Optimization!.

See @DrKungOptimizer 's new comprehensive preprint on AI Ethics

"Truss topology design under harmonic loads: Peak power minimization with semidefinite programming" by Shenyuan Ma et al, including IDA's Vyacheslav Kungurtsev @DrKungOptimizer, has been accepted for publication in Structural and Multidisciplinary Optimization!.

"State Encodings for GNN-based Lifted Planners" by R. Horcik, our @GustavSheer, V. Šimek, and T. Pevný, providing a comparative study of diverse state encodings for graph neural networks in classical planning, was accepted for oral presentation (top 20%) at AAAI25! @RealAAAI.

"A Stochastic-Gradient-Based Interior-Point Algorithm for Solving Smooth Bound-Constrained Optimization Problems" by Qi Wang, Frank E. Curtis, Daniel P. Robinson, and our V. Kungurtsev @DrKungOptimizer, has been accepted for publication in the SIAM Journal of Optimization!.

"Group Distributionally Robust Dataset Distillation with Risk Minimization", co-authored by our V. Kungurtsev, accepted for ICLR 2025! The paper uses Distributionally Robust Optimization techniques to improve generalization and subgroup coverage in Dataset Distillation techniques.

RT @ctu_cs: After the suspensful final this Saturday, Qminers Quant Hackathon #QQH2024 knows its winners! 🏆 The @FIT_CTU team won this spor….

RT @ctu_cs: Yesterday we hosted our annual department meeting. We heard from all research groups and administrators and learned about our a….

Towards Diverse Device Heterogeneous Federated Learning via Task Arithmetic Knowledge Integration by M Morafah, V Kungurtsev (IDA), H M Chang, C Chen, B Lin, presenting an algorithm for performing knowledge distillation across distinct computing devices accepted for NeurIPs 2024!.

"Unlocking the Potential of Federated Learning: The Symphony of Dataset Distillation via Deep Generative Latents" by Jia, Vahidian, Sun, Zhang, Vyacheslav Kungurtsev (of IDA), Gong, and Chen was accepted to the ECCV.

From @Aishine___ for today's last seminar before the summer break. enjoy the summer everyone!

Our IDA group representation at the recent "Void" event for curious @CVUTFEL students

ICAPS'24 Best Paper Runner-up Award goes to R. Horcik and our G. Sir (@GustavSheer) for their work on "Expressiveness of Graph Neural Networks in Planning Domains" - - find the paper at or the author(s) at @ICAPSConference in Banff!.

"Federated SGD with Local Asynchrony" by B. Chatterjee, our V. Kungurtsev, and D. Alistarh, studying the use of both parallel shared memory and distributed computation with asynchronous communication, was accepted for the IEEE Intl. Conference for Distributed Computing Systems.

"Efficient Dataset Distillation via Minimax Diffusion" with authors from Duke University and our V. Kungurtsev was accepted for CVPR. The paper harnessed the computational power of generative diffusions for synthesizing a rich and representative synthetic dataset for retraining.

and find the code repo at:

you can also watch a short video introduction:

"A Deep Learning Blueprint for Relational Databases" by L. Zahradnik, J. Neumann, and @GustavSheer, presented last week at #NeurIPS2023's @TrlWorkshop introduced a modular message-passing scheme for end-to-end deep learning from databases - see the paper!