User Portlet
Featured Contributor
Yaroslav Bulatov
Facebook AI
LOCATION: San Francisco
WEBSITE: Not indicated
INTERESTS IN JOBS & NETWORKING: Not indicated
ABOUT ME:
I'm a software engineer on Facebook PyTorch team, working on open-source for distributed neural network training.
I've used Mathematica since late 2000 to learn new things. My first post on MathGroup was in 2002, back from my grad school days when I worked on computational geometry. Ten years ago I actively blogged about graph theory and Mathematica, and participated in Mathematica stackoverflow community
- https://yaroslavvb.blogspot.com/2010
- http://mathematica-bits.blogspot.com/
- https://yaroslavvb.blogspot.com/2011
After that period I went to work at Google on large scale engineering problems at Google Maps and Google Brain team, involving machine learning and neural networks. Recently I've been using Mathematica again to understand how neural networks optimize: example
STAFF PICKS:
- High-dimensional sphere packing doesn't fit in a box
- How to count things in numerically stable way
- Neural networks, bird populations, complex analysis and DPR1 matrices
- Tensor networks, Einsum optimizers and the path towards autodiff
- Finding large graph cuts with SemidefiniteOptimization
- Inferring shape of optimization surface from shape of loss curve
- Obtaining Gaussian expectation formulas with einsums and TensorSimplify