Complex Numbers in PyTorch | PyTorch Developer Day 2020

preview_player
Показать описание
PyTorch is committed to providing best-in-class complex number support for scientific computing and AI research applications. In this talk, software engineer Anjali Chourdia discusses the state of complex functionality in PyTorch 1.7, including newly added linear algebra and spectral operations, and automatic differentiation support, as well as the plans for future releases.
Рекомендации по теме
Комментарии
Автор

Do PyTorch 1.7 Complex numbers support complex DNN construction? In other words, can the weights & biases & activation functions all be complex in the DNN and there is autograd for them?

henryzheng