Getting good GPU performance using Tensors in Matlab - experience from estimating a latent class non-convex model in Marketing/Economics. By Simon Yin
With the emergence of deep learning on GPUs and the future potential of quantum computing, user written code will need to evolve to express data parallelism in a compact way. In this short presentation, the use of tensors in Matlab will be explained and code for familiar scalar and vector operations will be expressed in terms of tensors (n-dimensional arrays). You will gain by the end of this session a simple set of coding patterns for accelerating your own numerical code.
Simon's talk will be followed by our monthly Hacky Hour where researchers (staff or students) from all disciplines can drop in to work on their research problems related to code, data, or digital tools in a friendly environment.