Abstract
Among the recent innovative technologies, memristor (memory-resistor) has attracted researchers attention as a fundamental computation element. It has been experimentally shown that memristive elements can emulate synaptic dynamics and are even capable of supporting spike timing dependent plasticity (STDP), an important adaptation rule that is gaining particular interest because of its simplicity and biological plausibility. The overall goal of this work is to provide a novel (theoretical) analog computing platform based on memristor devices and recurrent neural networks that exploits the memristor device physics to implement two variations of the backpropagation algorithm: recurrent backpropagation and equilibrium propagation. In the first learning technique, the use of memristor–based synaptic weights permits to propagate the error signals in the network by means of the nonlinear dynamics via an analog side network. This makes the processing non-digital and different from the current procedures. However, the necessity of a side analog network for the propagation of error derivatives makes this technique still highly biologically implausible. In order to solve this limitation, it is therefore proposed an alternative solution to the use of a side network by introducing a learning technique used for energy-based models: equilibrium propagation. Experimental results show that both approaches significantly outperform conventional architectures used for pattern reconstruction. Furthermore, due to the high suitability for VLSI implementation of the equilibrium propagation learning rule, additional results on the classification of the MNIST dataset are here reported.
Highlights
In the last few decades, the search of innovative computing platforms that could offer new, ultra-low power processing methods and architectures has intensified
Neuromorphic computing approaches aim to go beyond the state-of-the-art in conventional digital processing by exploiting complex dynamics and nonlinear phenomena emerging from the physics of nonvolatile memory devices (Chua, 1971; Strukov et al, 2008)
The network is trained by using two different generalizations of the backpropagation algorithm adapted to the continuous domain and energy-based models
Summary
In the last few decades, the search of innovative computing platforms that could offer new, ultra-low power processing methods and architectures has intensified. Neuromorphic computing approaches aim to go beyond the state-of-the-art in conventional digital processing by exploiting complex dynamics and nonlinear phenomena emerging from the physics of nonvolatile memory devices (e.g., memristors) (Chua, 1971; Strukov et al, 2008). The hallmark of this kind of devices is the peculiar analog signal storing capability that allows them to mimic the behavior of neural synapses. The processing is analog and different from current digital processors, and enhances computing speed and power efficiency for large sets of sensor data This has been achieved by combining memristor technology with advanced deep learning algorithms used to train neural networks.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
