Reliable edge machine learning hardware for scientific applications

Extreme data rate scientific experiments create massive amounts of data that require efficient ML edge processing. This leads to unique validation challenges for VLSI implementations of ML algorithms: enabling bit-accurate functional simulations for performance validation in experimental software frameworks, verifying those ML models are robust under extreme quantization and pruning, and enabling ultra-fine-grained model inspection for efficient fault tolerance. We discuss approaches to developing and validating reliable algorithms at the scientific edge under such strict latency, resource, power, and area requirements in extreme experimental environments. We study metrics for developing robust algorithms, present preliminary results and mitigation strategies, and conclude with an outlook of these and future directions of research towards the longer-term goal of developing autonomous scientific experimentation methods for accelerated scientific discovery.

I. Motivation

Ground-breaking science requires instruments that push sensing technology with increasing spatial and temporal resolution to explore nature at unprecedented scales and in extreme environments. This has led to a data generation explosion, with more and more data being generated in next-generation experiments. For example, particle physics experiments look for extremely rare collision events (one in a billion billion) that can answer fundamental questions about the fabric of space-time or the nature of dark matter. Alternatively, microscopy experiments take hundreds of thousands of images per second to understand material properties that can advance computing, quantum science, and basic energy research. There are many other applications in a wide range of domain sciences, including fusion, nuclear physics, neuroscience, and quantum computing, that can benefit from real-time, low-latency edge processing [1].