Talks and presentations

Fast Summation on a Sphere

February 28, 2023

Talk, SIAM CSE 23, Amsterdam, the Netherlands

Fast summation techniques have proven to be of great importance in a variety of fields. In this talk, I will present a new technique for performing fast summations on spheres, which is suitable for sums coming from spherical convolutions. I present applications of this technique to problems coming from atmospheric and oceanic modeling.

Communication Bounds for Convolutional Neural Networks

July 12, 2022

Talk, 2022 SIAM Annual Meeting, Pittsburgh, USA

Convolutional neural networks (CNNs) are important in a wide variety of machine learning tasks and applications, so optimizing their performance is essential. Moving words of data between levels of a memory hierarchy or between processors on a network is much more expensive than the cost of arithmetic, so minimizing communication is critical to optimizing performance. In this paper, we present new precise lower bounds on data movement for convolutions in both single-processor and parallel distributed memory models, as well as algorithms that outperform current implementations such as Im2Col. We obtain performance figures using GEMMINI, a machine learning accelerator, where our tiling provides improvements between 13% and 150% over a vendor supplied algorithm.

Communication Bounds for Convolutional Neural Networks

June 27, 2022

Talk, PASC '22: The Platform for Advanced Scientific Computing', Basel, Switzerland

Convolutional neural networks (CNNs) are important in a wide variety of machine learning tasks and applications, so optimizing their performance is essential. Moving words of data between levels of a memory hierarchy or between processors on a network is much more expensive than the cost of arithmetic, so minimizing communication is critical to optimizing performance. In this paper, we present new precise lower bounds on data movement for convolutions in both single-processor and parallel distributed memory models, as well as algorithms that outperform current implementations such as Im2Col. We obtain performance figures using GEMMINI, a machine learning accelerator, where our tiling provides improvements between 13% and 150% over a vendor supplied algorithm.