Anastasios Tefas

All-optical recurrent neural network with sigmoid activation function

2020

We demonstrate experimentally, the first all-optical recurrent-neuron with a sigmoid activation function and four WDM-inputs with 100psec pulses. The proposed neuron geared up a neural-network for financial prediction-tasks exhibiting an accuracy of 42.57% on FI-2010.

Programmable Photonic Neural Networks for advanced Machine Learning tasks

2023

Photonics holds the promise of reshaping Machine Learning and High-Performance Computing hardware landscape, stripping it of unnecessary signal conversion overhead, complying with strict power dissipation envelopes while unlocking unrivaled compute and bandwidth capacity.