Special Session: Neuromorphic hardware design and reliability from traditional CMOS to emerging technologies

Pavanello, Fabio; Vatajelu, Ioana; Bosio, Alberto; Vaerenbergh, Thomas van; Bienstman, Peter; Charbonnier, B.; Alessio, Carpegna,; Carlo, Stefano Di; Savino, Alessandro

doi:10.1109/vts56346.2023.10139932

Cited by 2 publications

(1 citation statement)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with electronics-based or other neuromorphic computing platforms [15,16,17], photonics offers unique advantages in terms of parallelism, energy efficiency, latency and bandwidth of interconnects [18,19,13]. These are particularly relevant for the development of large-scale hardware ANNs, which comprise a huge number of parallel weighted connections (synapses).…”

Section: Introductionmentioning

confidence: 99%

Emergent self-adaptation in an integrated photonic neural network for backpropagation-free learning

Lugnan,

Aggarwal,

Brückerhoff-Plückelmann

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Plastic self-adaptation, nonlinear recurrent dynamics and multi-scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt and process information similarly to the way biological brains do. In this work, we experimentally demonstrate these properties occurring in arrays of photonic neurons. Importantly, this is realised autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation-free training algorithm based on simple logistic regression, we are able to achieve a performance of 98.2% on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase-change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed.

show abstract

Section: Introductionmentioning

confidence: 99%

Emergent self-adaptation in an integrated photonic neural network for backpropagation-free learning

Lugnan,

Aggarwal,

Brückerhoff-Plückelmann

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Exploring the potential of self-pulsing optical microresonators for spiking neural networks and sensing

Biasi,

Lugnan,

Micheli

et al. 2024

Commun Phys

View full text Add to dashboard Cite

Photonic platforms are promising for implementing neuromorphic hardware due to their high processing speed, low power consumption, and ability to perform parallel processing. A ubiquitous device in integrated photonics, which has been extensively employed for the realization of optical neuromorphic hardware, is the microresonator. The ability of CMOS-compatible silicon microring resonators to store energy enhances the nonlinear interaction between light and matter, enabling energy efficient nonlinearity, fading memory and the generation of spikes via self-pulsing. In the self-pulsing regime, a constant input signal can be transformed into a time-dependent signal based on pulse sequences. Previous research has shown that self-pulsing enables the microresonator to function as an energy-efficient artificial spiking neuron. Here, we extend the experimental study of single and coupled microresonators in the self-pulsing regime to confirm their potential as building blocks for scalable photonic spiking neural networks. Furthermore, we demonstrate their potential for introducing all-optical long-term memory and event detection capabilities into integrated photonic neural networks. In particular, we show all-optical long-term memory up to at least 10 μ s and detection of input spike rates, which is encoded into different stable self-pulsing dynamics.

show abstract

Special Session: Neuromorphic hardware design and reliability from traditional CMOS to emerging technologies

Cited by 2 publications

References 72 publications

Emergent self-adaptation in an integrated photonic neural network for backpropagation-free learning

Emergent self-adaptation in an integrated photonic neural network for backpropagation-free learning

Exploring the potential of self-pulsing optical microresonators for spiking neural networks and sensing

Contact Info

Product

Resources

About