Model conversion method, processing chip and electronic equipment

Abstract

The embodiment of the invention provides a model conversion method, a processing chip and electronic equipment, and relates to the technical field of computers. The model conversion method comprises the steps: obtaining branch node information and leaf node information of a decision tree, wherein the branch node information comprises a plurality of branch nodes and decision conditions of the branch nodes, and the leaf node information comprises a plurality of leaf nodes and sample types to which the leaf nodes belong; based on the types of an input sample and an output sample of the decision tree, the branch node information and the leaf node information, constructing a structure of a spiking neural network corresponding to the decision tree; and setting weight distribution between adjacent layers in the spiking neural network based on the decision condition of each branch node and the sample type to which each leaf node belongs. According to the invention, equivalent transformation from the decision tree to the spiking neural network is realized.

Description

technical field [0001] The invention relates to the technical field of computers, in particular to a model conversion method, a processing chip and electronic equipment. Background technique [0002] Spiking Neural Network (SNN for short) has attracted the attention of academia and industry in recent years due to its low power consumption and characteristics closer to the human brain. In a spiking neural network, the axon is the unit that receives pulses, and the neuron is the unit that sends pulses. A neuron is connected to multiple axons through dendrites, and the connection point between dendrites and axons is called a synapse. After the axon receives the pulse, all dendrites that have synaptic connections with this axon will receive the pulse, which in turn affects the downstream neurons of the dendrite. The neuron sums the pulses from multiple axons and accumulates the previous membrane voltage, sending a pulse downstream if the value exceeds a threshold. The pulse ne...

AI Technical Summary

Problems solved by technology

[0003] The development of the second generation artificial neural network (ANN for short) has already been very mature, but the way of pure digital encoding information of ANN artificial neural network has resulted in many ANN models, such as convolutional neural network (CNN), recurrent neural network (RNN) There is a large loss of precision after the model is converted into an SNN model, and the prediction accuracy is not as good as the original model

However, if you want to convert the random forest model to a spiking neural network model to reduce the computing power consumption and power consumption of running the random forest model, there will also be a problem of loss of accuracy

Images