Scholar
Soham De
Google Scholar ID: lHf55pF3KVQC
DeepMind
Citations & Impact
Citations
6,552
H-index
28
i10-index
33
Publications
20
Co-authors
58
list available
Publications
Resume (English only)
Academic Achievements
- Published several papers including 'Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models' (Preprint), 'Resurrecting Recurrent Neural Networks for Long Sequences' (ICML 2023), and 'Unlocking High-Accuracy Differentially Private Image Classification through Scale' (Preprint). Delivered multiple talks at venues such as ICML workshops, Google Brain, and others.
Research Experience
- Currently a research scientist at DeepMind, focusing on better understanding and improving large-scale deep learning. During his PhD, he worked on fast stochastic optimization algorithms and collaborated with Michele Gelfand on game-theoretic models of the evolution of human behavioral norms. He has previously interned at Toyota Technological Institute at Chicago (TTIC), IBM Research Almaden, and DeepMind.
Education
- Completed PhD in 2018 at the University of Maryland, advised by Dana Nau and Tom Goldstein. Completed undergraduate degree from Jadavpur University, Kolkata, India in 2013.
Background
- Research interests include optimization and learning dynamics, efficient training and inference, scaling principles, and the robustness and privacy of deep learning models.
