Designing Reactive AI With Spiking Neural Nets

Wolfram-like attention framing meets spiking networks: event-triggered, energy-thrifty AI that “wakes” to stimuli.
GitHub

Accelerating Training with Neuron Interaction and Nowcasting Networks

We introduce the NiNo model predicting future (nowcasting) parameters by learning neuron interaction in vision and language tasks. We feed c (c=5 by default) past parameter states as input to NiNo and ...
IEEE

Efficient Deep Neural Network Training With a Novel 5.3-Bit Block Floating Point Data Format

Abstract: Low-bit-width data formats offer a promising solution for enhancing the energy efficiency of Deep Neural Network (DNN) training accelerators. In this work, we introduce a novel 5.3-bit data ...
GitHub

Coupling LSTM Neural Networks and State-Space Models through Analytically Tractable Inference

This repo contains the matlab codes to reproduce the results for the paper: Vuong, Nguyen & Goulet (2024), Coupling LSTM Neural Networks and State-Space Models through Analytically Tractable Inference ...
IEEE

Layer-Wise Training of Graph Neural Networks With Self-Supervised Learning

Abstract: Training graph neural networks (GNNs) on large graphs is challenging due to both the high memory and computational costs of end-to-end training and the scarcity of detailed node-level ...