Spiking neural networks compensate for weight drift in organic neuromorphic device networks

Abstract

Organic neuromorphic devices can accelerate neural networks and integrate with biological systems. Devices based on the biocompatible and conductive polymer PEDOT:PSS are fast, require low amounts of energy and perform well in crossbar simulations. However, parasitic electrochemical reactions lead to self-discharge and the fading of learned conductance states over time. This limits a neural network's operating time and requires complex compensation mechanisms. Spiking neural networks take inspiration from biology to implement local and always-on learning. We show that these spiking neural networks can function on organic neuromorphic hardware and compensate for self-discharge by continuously relearning and reinforcing forgotten states. In this work, we use a high-resolution charge transport model to describe the behavior of organic neuromorphic devices and create a computationally-efficient surrogate model. By integrating the surrogate model into a Brian 2 simulation, we can describe the behavior of spiking neural networks on organic neuromorphic hardware. A biologically-plausible two-layer network for recognizing 28x28 pixel MNIST images is trained and observed during self-discharge. The network achieves, for its size, competitive recognition results up to 82.5%. Building the network with forgetful devices yields superior accuracy during training with 84.5% compared to ideal devices. However, the trained networks without active spike-timing-dependent plasticity quickly lose their predictive performance. We show that online learning can keep the performance at a steady level close to the initial accuracy, even for idle rates of up to 90%. This performance is maintained when the output neuron's labels are not revalidated for up to 24 hours. These findings reconfirm the potential of organic neuromorphic devices for brain-inspired computing. Their biocompatibility and the demonstrated adaptability to spiking neural networks open the path toward close integration with multi-electrode arrays, drug-delivery devices, and other bio-interfacing systems as either full organic or hybrid organic-inorganic systems.

Editorial: Focus on organic materials, bio-interfacing and processing in neuromorphic computing and artificial sensory applications

“…Using a similar network of organic ECRAMs, which is traditionally implemented for in-memory computing, the same authors show in Felder et al [2] that it is also possible to run spiking neural networks. They demonstrate that their non-ideal devices that 'forget' their state, are actually 'reminded' by the spikes, and thus keep the high classification accuracy.…”