Naive Bayes classification method, engine and system based on resistive random access memory array

Abstract

The invention discloses a naive Bayes classification method, engine and system based on a resistive random access memory array. The method belongs to the field of storage and calculation fusion. The method comprises the steps that a monotonically decreasing function f is used for transforming naive Bayes to convert continuous multiplication into dot multiplication, the prior probability and the conditional probability are respectively changed into sums to distribute a bit line for each category, and on each bit line, a storage unit is correspondingly distributed for each prior probability andeach conditional probability; setting the conductance of the storage unit to be related to the corresponding prior probability or conditional probability; According to the value of each attribute in the to-be-classified instance, applying a voltage to the corresponding word line, Meanwhile, applying a voltage to the word line corresponding to the prior probability, so that the current of the storage unit is correspondingly equal to or acquired from the current on each bit line, and the category corresponding to the minimum current is determined as a final classification result. The delay and power consumption of the naive Bayesian classification method can be reduced.

Description

technical field [0001] The invention belongs to the field of fusion of storage and computing, and more specifically relates to a naive Bayesian classification method and an engine based on a resistive storage array. Background technique [0002] Resistive Random-Access Memory (ReRAM) is a new type of non-volatile passive two-terminal memory device. It can be composed of specific resistive materials (for example, some metal oxides), and the stored information can be represented by the level of cell resistance. Generally, a high-impedance state of a cell represents a logic 0, and a low-impedance state of a cell represents a logic 1. Further subdividing the unit resistance value can be used to represent multi-bit values. As shown in Figure 1(a), it is the conductive filament model of the ReRAM unit. The conductive filament connects the upper and lower poles and presents the lowest resistance value, and vice versa, it presents the highest resistance value. Applying forward an...

AI Technical Summary

Problems solved by technology

[0009] At present, the naive Bayesian algorithm is mostly based on software implementation, and the storage and computing are separated under the traditional von Neumann architecture, which makes the memory frequently accessed when using the naive Bayesian algorithm for classification, resulting in data between the CPU and the memory. Migrate frequently, resulting in slow solution speed, high latency, and high power consumption

Images