BRAInLab

BRAIn Lab × MBZUAI — 5 years of collaboration 🔥 Together, we’ve prepared a survey on Byzantine attacks — link below 👇 mbzuai.ac.ae/news/the-searc… Despite a challenging period in Abu Dhabi, the work continues. Supporting our colleagues 🙏

AAAI-26 — done ✅ Emotions overflowing, great scientific connections, and tons of ideas. Next up — ICLR 👀✨

To get more details👇🏻 📄 Paper: arxiv.org/pdf/2511.04569 🖇️ References: [1] arxiv.org/pdf/1905.11261 [2] arxiv.org/pdf/1906.01133 [3] proceedings.neurips.cc/paper_files/pa…

As a result, these adaptive step sizes behave exactly like the tuned ones — or slightly better

Adaptive step sizes without knowing the smoothness constant? That’s real! Compatible with ANY variance-reduction method — finite-sum, distributed, coordinate sketches included 📜 Unified Theory of Adaptive Variance Reduction From pure SGD to advances like variance reduction

Ready to dive into the details or try it yourself? 🚀 📄 Paper: arxiv.org/pdf/2508.16734 💻 Code: github.com/brain-lab-rese…

We tested ALSO across diverse setups with real-world data heterogeneity, including: • Class-imbalanced datasets • Tabular DL • Robust & distributed training • Split learning In all scenarios, ALSO showed superior performance over both standard approaches and DRO baselines

Struggling with data heterogeneity in your deep learning projects? We introduce ALSO, a new, practical Distributionally Robust Optimizer (DRO)🚀 📜 Paper thread for "Aligning Distributionally Robust Optimization with Practical Deep Learning Needs"