Abstract: Provided are a memory device and a method of forming the same. The memory device includes: a selector; a magnetic tunnel junction (MTJ) structure, disposed on the selector; a spin orbit torque (SOT) layer, disposed between the selector and the MTJ structure, wherein the SOT layer has a sidewall aligned with a sidewall of the selector; a transistor, wherein the transistor has a drain electrically coupled to the MTJ structure; a word line, electrically coupled to a gate of the transistor; a bit line, electrically coupled to the SOT layer; a first source line, electrically coupled to a source of the transistor; and a second source line, electrically coupled to the selector, wherein the transistor is configured to control a write signal flowing between the bit line and the second source line, and control a read signal flowing between the bit line and the first source line.

Abstract: Provided is a distributed insulation detection device for a multi-stage DC system. A basic insulation combination module is configured to detect a ground insulation fault of the multi-stage DC system, a sampling module is configured to collect voltage data and/or leakage current data of the multi-stage DC system, to transmit the collected voltage data and/or the leakage current data to an intelligent control module for data processing, to control a resistance value of the basic insulation combination module and a resistance value of an intelligent resistance switching network module, so as to adjust a total balance resistance of the distributed insulation detection device for the multi-stage DC system.

Abstract: A neural network training method, apparatus, a storage medium, and a computer device are provided. The method includes: obtaining a training sample set, each training sample including a standard label; inputting the each training sample into a neural network model including n attention networks, the n attention networks respectively mapping the each training sample to n subspaces, each of the n subspaces including a query vector sequence, a key vector sequence, and a value vector sequence; calculating a space difference degree between the n subspaces by using the neural network model; calculating an output similarity degree according to an output of the neural network model and the standard label corresponding to the each training sample; and adjusting a model parameter of the neural network model according to the space difference degree and the output similarity degree until a convergence condition is satisfied to obtain a target neural network model.