A hardware-aware liquid state machine network generation method and system
A liquid state, machine network technology, applied in the direction of biological neural network model, stochastic CAD, climate sustainability, etc., can solve the problem of not combining the LSM network structure with the brain-like processor structure, reducing the amount of communication, etc. Communication overhead, the effect of good classification accuracy
Active Publication Date: 2022-06-07
NAT UNIV OF DEFENSE TECH
View PDF8 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, this mapping method in the prior art only reduces a small amount of communication, and does not fundamentally combine the structure of the LSM network with the structure of the brain-like processor
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0041] like figure 2As shown, the hardware-aware liquid state machine network generation method in this embodiment includes:
[0042] 1) Initialization stage: On the premise that the LSM neurons of each core of the brain-like processor do not exceed the preset threshold, a group of unconnected LSM neurons are randomly mapped to each core of the brain-like processor; The positions of the cores assigned by the LSM neurons in the brain-like processor calculate the distance between any two LSM neurons, forming a distance matrix; see figure 2 , where the distance between any two LSM neurons is denoted as D i,j , forming a distance matrix and denoted as Distance Matrix(D);
[0043] 2) LSM parameter search stage: Generate a new LSM network structure by iteratively adjusting the parameter values of the control parameters λ and C of the LSM network, and calculate any two LSMs of the LSM network according to the input distance matrix through the new LSM network structure. The pro...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a method and system for generating a hardware-aware liquid state machine network. The method of the invention includes randomly mapping LSM neurons to a brain-like processor and generating a distance matrix; adjusting the control parameters λ, The parameter value of C guides the generation of a new LSM network structure, and calculates the probability of connection between any two LSM neurons for the input distance matrix to obtain the connection matrix, repeats the above iterative adjustment of the LSM network process until the termination condition is reached; the final Excellent LSM network structure output. The present invention can add distance information when generating the LSM network structure through initialization, so that the communication between neurons can be kept in the core as much as possible, greatly reducing the communication overhead between the cores, and using a heuristic algorithm to search for different LSM network structures, The generated LSM network can not only greatly reduce the communication overhead in the NoC, but also ensure good classification accuracy.
Description
technical field [0001] The invention relates to an automatic generation technology of a liquid state machine (LSM), in particular to a method and a system for generating a network of a hardware-aware liquid state machine. Background technique [0002] Liquid state machine model (liquid state machine, LSM network for short) is a kind of spiking neural network with loopback (Spiking Neural Network, SNN), due to its powerful computing power, biological rationality, simple structure and relatively It has attracted the interest of many researchers due to its low training complexity. Compared with other neural networks, it can perform image classification, speech recognition, etc. with little overhead. The structure of LSM is as figure 1 As shown, it mainly consists of three parts: input layer, liquid layer and output layer. The input layer is responsible for discrete spike inputs, and its neurons are connected to neurons in the liquid layer. The liquid layer is composed of re...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F30/25G06F30/27G06N3/04G06F111/08
CPCG06F30/25G06F30/27G06F2111/08G06N3/045Y02D10/00
Inventor 王蕾王世英曲连华康子扬李石明张剑锋刘威张英潘国腾苏金树
Owner NAT UNIV OF DEFENSE TECH