Gusciora

About

Gusciora’s research sits at the intersection of neural networks and high-performance computing, with a focus on accelerating machine learning through parallel architectures. Their most influential contribution, the 1988 paper “Neural network simulation at Warp speed: how we got 17 million connections per second” (161 citations), introduced a fast back-propagation algorithm designed for a linear array of processors. This work demonstrated a groundbreaking implementation on the Warp machine—a ten-processor, programmable systolic array computer—achieving 17 million connections per second, a speed that set a new benchmark for its era. By systematically comparing their results with back-propagation implementations on other machines, Gusciora provided a clear, quantitative proof that specialized parallel hardware could dramatically accelerate neural network training. This achievement not only advanced the practical feasibility of larger-scale neural simulations but also influenced subsequent research in hardware-software co-design for AI. Gusciora’s work remains a foundational reference for researchers exploring efficient, high-speed neural network implementations on parallel systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Neural network simulation at Warp speed: how we got 17 million connections per second

   161 citations · 1988

Key Collaborators

Contact & Links