Matrices and tensors - named, visualized
In data science, we often work with numeric arrays: for signals, images, accounting data - input, output, everything. When we use Pandas (vs raw NumPy) we have named dimensions (rows and columns) plus an easy way to plot the numerical values.
In this talk, I will show how to go beyond that - how to name dimensions in 3 and more dimensional arrays for deep learning (with Named Tensors in PyTorch), how to visualize advanced operations in a simple way (using tensor diagrams) and how to plot complex numbers (for quantum computing).
I will give examples from two open-source projects I develop: https://github.com/Quantum-Game/bra-ket-vue and https://github.com/stared/pytorch-named-dims.
AI Researcher / Game Dev
ECC Games / Quantum Flytrap
A deep learning specialist with a Ph.D. in quantum physics (from ICFO, Castelldefels).
He works on AI for content design and physics engine optimization in ECC Games, and develops Quantum Game with Photons - an open-source in-browser game with real quantum mechanics.
Piotr enjoys explaining complicated things in simple ways, ideally with interactive data visualizations. He develops livelossplot - a Python package for visualizing the training process in Jupyter Notebook.