Non-spike timing-dependent plasticity learning mechanism for memristive neural networks
- PDF / 6,303,501 Bytes
- 12 Pages / 595.224 x 790.955 pts Page_size
- 41 Downloads / 183 Views
Non-spike timing-dependent plasticity learning mechanism for memristive neural networks Zhiri Tang1,2 · Yanhua Chen3 · Zhihua Wang2 · Ruihan Hu4 · Edmond Q. Wu5 Accepted: 27 September 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Memristive neural networks (MNNs) attract the attention of many researchers because memristor can mimic the learning mechanism of biologic neuron, spike timing-dependent plasticity (STDP). While STDP brings huge potentials on many applications for memristive neural networks, it also gives complex calculation process for hardware implement. In this work, a non-STDP learning mechanism is proposed, which is implemented in two common frameworks including feedforward neural network and crossbar. The non-STDP learning mechanism relies on the linear relationship between the value of memristor and area of input spikes, which gives the proposed method a simple calculation process and better hardware compatibility. Experimental results show that the non-STDP learning mechanism can help to achieve good hardware performance in both feedforward neural network and crossbar frameworks. Compared with STDP based memristive neural networks, the proposed method can save 2.19%-24.4% hardware resource (ALMs) and improve 1.56-12.25 MHz processing speed under a set of different network scales. In future, some other complex memristor models with non-STDP learning mechanism should be taken into consideration, which will give more room for practical applications of memristive neural networks. Keywords Spike timing-dependent plasticity · Memristive neural networks · Feedforward neural networks · Crossbar · Hardware performance
1 Introduction 1.1 Motivation After memristor was proposed and proved by Leon Chua [5] and first made by HP Lab [27], memristive characteristics in mimicking learning rule of biologic neuron such as spike timing-dependent plasticity (STDP) [42] have attracted attention from many researchers. During the past few years, many works on STDP based memristive neural networks (SMNN) have emerged. S-MNN gives more inspirations on neuromorphic computing and helps to explore the applications of new memristor devices. Meanwhile, STDP brings complex calculation and hardware implement into the entire systems. The main goals and advantages of neuromorphic computing are low power, high speed, high throughput, few
Ruihan Hu
[email protected]
Extended author information available on the last page of the article.
hardware occupancy, and small chip area [4]. So it has a big dilemma when using STDP as the main learning mechanism of systems. Hence, a new learning mechanism needs to be put forward, which uses spikes as transmission signals and can also give higher hardware compatibility compared with traditional STDP based methods.
1.2 Related works In recent years, memristive neural networks (MNNs) [37, 45] have become one of the hottest topics in neuromorphic computing [4] due to memristor’s potential on descriptions of biologic synapse’s characteristics, spike timing-
Data Loading...
