Machine Learning, particularly Deep Learning, algorithms are being increasingly used to approximate solutions of partial differential equations (PDEs). We survey recent results on different aspects of deep learning in the context of PDEs namely, 1) Supervised learning for high-dimensional parametrized PDEs 2) Operator learning for approximating infinite-dimensional operators which arise in PDEs and 3) Physics informed Neural Networks for approximating both forward and inverse problems for PDEs. We will highlight open questions in the analysis of deep learning algorithms for PDEs.

The video of this talk is available on the IISc Math Department channel.