DIKU Bits: Differentiable Programming & Biology

Foredrag — Deep learning frameworks such as PyTorch, and JAX, and their approach to autodifferentiation, have spawned a revolution in the creativity of neural network architectural design. In this DIKUBit, Julius will explain how the idea of differentiable programming enables novel approaches in biology. He'll show that we can improve our understanding of how microbial cells sense their environment, why leaves look the way they do, and how we can improve algorithmic organism tracking using physical simulations.