Choudhary A Artificial Intelligence For Science A Deep Learning Revolution 2023

Textbook in PDF format

This unique collection introduces AI, Machine Learning (ML), and deep neural network technologies leading to scientific discovery from the datasets generated both by supercomputer simulation and by modern experimental facilities.
Huge quantities of experimental data come from many sources ? telescopes, satellites, gene sequencers, accelerators, and electron microscopes, including international facilities such as the Large Hadron Collider (LHC) at CERN in Geneva and the ITER Tokamak in France. These sources generate many petabytes moving to exabytes of data per year. Extracting scientific insights from these data is a major challenge for scientists, for whom the latest AI developments will be essential.
To work with Deep Learning methods, practitioners rely on Deep Learning frameworks — a specialized software which allows users to design and train Artificial Neural Networks (ANNs). Popular titles include PyTorch, TensorFlow, and MXNet among others. In summary, the frameworks allow users to (a) construct an ANN from basic tensor operations and a predefined set of basic building blocks (convolutional kernels, dense layers, recurrent units, various activation functions, etc.), (b) compute derivatives of the loss function for back-propagation, (c) execute forward and backward passes efficiently on a given hardware, and (d) provide additional helper functions for distributed training, data handling, etc. Modern Deep Learning frameworks are extremely complicated software products and are typically designed in a hierarchical and modular fashion.
The timely handbook benefits professionals, researchers, academics, and students in all fields of science and engineering as well as AI, ML, and neural networks. Further, the vision evident in this book inspires all those who influence or are influenced by scientific progress