Ecosyste.ms sponsors
An open API service aggregating public data about GitHub Sponsors.
An open API service aggregating public data about GitHub Sponsors.
Research MSc at @mila-iqia. VRS @VITA-Group, @humansensinglab and Founder @landskape-ai. ボイド
Funding Links: https://github.com/sponsors/digantamisra98
Hi there!
I'm Diganta Misra, founder of a research group Landskape and Machine Learning Engineer at Weights & Biases. I'm also an incoming MSc in CS (Machine Learning specialization) at MILA, Montreal affiliated with UdeM and a Visiting Research Scholar at VITA, UT Austin. I mostly focus on Abstract Algebra, Computer Vision, Mean Field Theory, Continual Learning, Convex Optimization, Deep Learning Theory and Non-Linear Dynamics.
As an open-source and reproducibility enthusiast, I aim to make current research in deep learning more transparent, reproducible and accessible. Landskape is a non-profit research org that I founded to provide a platform to students and like-minded professionals to explore research ideas of common interests for understanding deep neural networks.
Running an organization as well as working on reproducibility comes with financial and compute requirements. With GitHub sponsors, it will allow me to continue with my efforts in this domain and also fund my research.
My current research interest is to make sparse models cool and also have large scale robust models. As of right now, I am working on theory of robustness in deep learning.
I also want to make a podcast where I interview less well-known authors whose works are leaving an imprint in various domains of deep learning and debunk their papers along with them from a theoretical, applied and reproducibility perspective.
Official Repository for "Mish: A Self Regularized Non-Monotonic Neural Activation Function" [BMVC 2020]
Language: Jupyter Notebook - Stars: 1292Python package containing all custom layers used in Neural Networks (Compatible with PyTorch, TensorFlow and MegEngine)
Language: Python - Stars: 137Unofficial PyTorch Implementation of EvoNorm
Language: Python - Stars: 121Unofficial Implementation of ECANets (CVPR 2020) for the Reproducibility Challenge 2020.
Language: Python - Stars: 32