Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses

Abstract

In an increasingly data-rich world the need for developing computing systems that cannot only process, but ideally also interpret big data is becoming continuously more pressing. Brain-inspired concepts have shown great promise towards addressing this need. Here we demonstrate unsupervised learning in a probabilistic neural network that utilizes metal-oxide memristive devices as multi-state synapses. Our approach can be exploited for processing unlabelled data and can adapt to time-varying clusters that underlie incoming data by supporting the capability of reversible unsupervised learning. The potential of this work is showcased through the demonstration of successful learning in the presence of corrupted input data and probabilistic neurons, thus paving the way towards robust big-data processors.

A Data-Driven Verilog-A ReRAM Model

“…Fig. 7 reveals the switching rate surface reproduced by the proposed model expressions (8) and the parameter values (9), (10) along with the switching rate measurements that resulted by the application of the optimizer routine on the DUT. The particularly good matching of the proposed expression (8) versus experimental results corroborates the assumptions upon which our model was structured.…”

“…What we propose is the straightforward integration of the state variable function (8) with the use of the built in VA time-domain integration operator "idt()". Numerical integration may produce errors in sensitive models with extremely long time constants but this operation is far more preferable than time-domain differentiation [8].…”

“…Numerical integration may produce errors in sensitive models with extremely long time constants but this operation is far more preferable than time-domain differentiation [8]. The integral present in (8) corresponds to a recursive relationship.…”

“…Such devices are becoming more and more accessible to researches, and it is now clear that each implementation features properties that render them suitable for different applications. There are memristors that have been reported to switch quickly and in a probabilistic fashion [6], while others can have their resistive state (RS) shifted in small steps (analogue mode) [7] which is ideal for synaptic learning [8].…”

Unsupervised Learning Using Charge-Trap Transistors

“…For each presentation of a pattern, the neuron with the largest output fires and claims the pattern, and only the 9 synaptic weights associated with this neuron are updated with a WTA rule [19]. Specifically, only when output neuron j has the largest output and fires (wins) are G OFFi,j (i = 1−9) updated: ΔG OFFi,j is increased by detrapping pulses if the input neuron i also fires or decreased by trapping pulses if the input neuron i does not fire.…”