TENNs-PLEIADES: Building Temporal Kernels with Orthogonal Polynomials

We introduce a neural network named PLEIADES (PoLynomial Expansion InAdaptive Distributed Event-based Systems), belonging to the TENNs (TemporalNeural Networks) architecture. We focus on interfacing these networks withevent-based data to perform online spatiotemporal classification and detectionwith low latency. By virtue of using structured temporal kernels andevent-based data, we have the freedom to vary the sample rate of the data alongwith the discretization step-size of the network without additional finetuning.We experimented with three event-based benchmarks and obtained state-of-the-artresults on all three by large margins with significantly smaller memory andcompute costs. We achieved: 1) 99.59% accuracy with 192K parameters on theDVS128 hand gesture recognition dataset and 100% with a small additional outputfilter; 2) 99.58% test accuracy with 277K parameters on the AIS 2024 eyetracking challenge; and 3) 0.556 mAP with 576k parameters on the PROPHESEE 1Megapixel Automotive Detection Dataset.