Neuromorphic Photonic Computing using Frequency Multiplexing
Serge Massar | Université libre de Bruxelles
Neuromorphic computing aims to implement computations in ways very loosely inspired by how the brain works. Such computation can be implemented in different platforms, including optics. We describe photonic implementations in the frequency domain. Individual “neurons” are encoded as the amplitudes of the spectral lines of a frequency comb. We describe realizations of Extreme Learning Machines and reservoir computers, as well as deep reservoir computers. These implementations use standard telecommunication components. Mixing between comb lines is realized either by electro-optic modulation or nonlinear 4 wave mixing. All optical readout is realized using a programmable spectral filter. This approach demonstrates the parallelism offered by light since multiple comb lines are processed simultaneously by the same optical components.
About the speaker
Serge Massar obtained his PhD from Université libre de Bruxelles (ULB) in 1995. After postdocs in Tel Aviv and Utrecht, he joined ULB in 1998 as Research Associate of the FRS-FNRS. Currently he is full Professor at ULB. His research interests include quantum information, quantum and nonlinear optics, photonic neuromorphic computing. He has co-authored more than 200 scientific articles and conference proceedings. His work has been recognized by several prizes, including the Gödel prize in 2023. Since 2021 he is member of the Royal Academy of Science, Letters and Fine Arts of Belgium.
Location
QNC 1201