Time-Delay Reservoir Computing with Optically Injected Low-Threshold Semiconductor Lasers
In this talk I will present my work on time-delay reservoir computing based on ultra-low-threshold semiconductor lasers, a photonic computing approach that can potentially offers a promising route toward more energy-efficient processing of temporal information. In this context, such devices are particularly attractive due to their reduced operating current and their strong sensitivity to optical perturbations. I will focus on how delayed feedback introduces memory into the photonic reservoir, while optical injection and frequency detuning provide control over the laser dynamics. These parameters will enable access to different operational regimes, including injection locking and transition regions characterized by enhanced nonlinear responses. The Santa Fe time-series prediction task will be used as a benchmark task for evaluating the computational efficiency. The normalized mean squared error (NMSE) will be used to evaluate one-step-ahead prediction performance. To better understand the underlying dynamics, I will compare the normalized optical input and reservoir output signals, I will also show a detuning-dependent cross-correlation analysis to assess the temporal information processing capabilities. In addition, virtual-node correlation analysis will be used to quantify the diversity and redundancy of reservoir states.
This Annual PhD student seminar will be broadcasted in the following zoom link: https://us06web.zoom.us/j/89466064429?pwd=po9p99eAEYVPaNI8xIIGoOIz0hOqaF.1
Detalles de contacto:
Miguel C. Soriano Contact form