The present invention provides a method and system for processing data from an event, such as a neurological event. When a neurological event occurs, a spike in a neural waveform is generated. The spike can be detected and used to determine information about the neurological event. The method uses data values from a resistive switching component capable of undergoing a resistive state change when a voltage is applied to it. The data values represent a sequence of resistive state changes of the resistive switching component which correspond to the neurological event. The method further comprises processing the received data values to identify a resistive state change corresponding to the neurological event and to obtain information about the neurological event. Thus, a method and system for processing neural spikes is provided.

The present invention provides a method and system for processing data from an event, such as a neurological event. When a neurological event occurs, a spike in a neural waveform is generated. The spike can be detected and used to determine information about the neurological event. The method uses data values from a resistive switching component capable of undergoing a resistive state change when a voltage is applied to it. The data values represent a sequence of resistive state changes of the resistive switching component which correspond to the neurological event. The method further comprises processing the received data values to identify a resistive state change corresponding to the neurological event and to obtain information about the neurological event. Thus, a method and system for processing neural spikes is provided.

A semiconductor device includes a ferroelectric field-effect transistor (FeFET), wherein the FeFET includes a substrate; a source region in the substrate; a drain region in the substrate; and a gate structure over the substrate and between the source region and the drain region. The gate structure includes a gate dielectric layer over the substrate; a ferroelectric film over the gate dielectric layer; and a gate electrode over the ferroelectric film.

A semiconductor device capable of holding analog data is provided. Two holding circuits, two bootstrap circuits, and one source follower circuit are formed with use of four transistors and two capacitors. A memory node is provided in each of the two holding circuits; a data potential is written to one of the memory nodes and a reference potential is written to the other of the memory nodes. At the time of data reading, the potential of the one memory node is increased in one of the bootstrap circuits, and the potential of the other memory node is increased in the other of the bootstrap circuits. A potential difference between the two memory nodes is output by the source follower circuit. With use of the source follower circuit, the output impedance can be reduced.

Examples include a method of programming resistive random access memory (RRAM) array for analog computations. In some examples, a selected RRAM cell of the RRAM array may be programmed with a selected target conductance and a programmed conductance error of the selected RRAM cell may be determined. A neighboring RRAM cell may be programmed with an error corrected target conductance that is a function of a neighboring target conductance and the programmed conductance error of the selected RRAM cell. The neighboring RRAM cell may be in a same row or a same column as the selected RRAM cell. The selected RRAM cell and neighboring RRAM cell are programmed such that the RRAM array is programmed for an analog computation.

A system of processing a task based on information of frequently used algorithms learned through a memory unit includes a first memory, a second memory, a processor, and a reading unit. The processor processes a first type of task using a first algorithm, and writes to a first memory cell of the second memory. The second memory including first and second memory cells each having a charge storage element. The first and second memory cells correspond to the first and second algorithms, respectively. The reading unit senses a first voltage stored in the first memory cell and a second voltage stored in the second memory cell, and provides information of frequently used algorithms to the processing device based on the sensed first and second voltages.

An embodiment of a semiconductor apparatus may include technology to convert an analog voltage level of a memory cell of a multi-level memory to a multi-bit digital value, and determine a single-bit value of the memory cell based on the multi-bit digital value. Some embodiments may also include technology to track a temporal history of accesses to the memory cell for a duration in excess of ten seconds, and determine the single-bit value of the memory cell based on the multi-bit digital value and the temporal history. Other embodiments are disclosed and claimed.

A method of manufacturing an electronic device including a semiconductor memory is provided. The method may include forming a material layer for forming a variable resistance element over a substrate, forming a metal layer over the material layer, forming a mask pattern over the metal layer, forming a metal layer pattern by etching the metal layer using the mask pattern as an etch barrier, performing a surface treatment on the metal layer pattern, and etching the material layer using the metal layer pattern and the metal compound layer as an etch barrier to form a variable resistance element having an external side aligned with an external side of the metal compound layer. An external part of the metal layer pattern may be transformed into a metal compound layer. The metal compound layer may have a low etch rate as an etch barrier.

A static random-access memory (SRAM) system includes SRAM cells configured to perform exclusive NOR operations between a stored binary weight value and a provided binary input value. In some embodiments, SRAM cells are configured to perform exclusive NOR operations between a stored binary weight value and a provided ternary input value. The SRAM cells are suitable for the efficient implementation of emerging deep neural network technologies such as binary neural networks and XNOR neural networks.

A sample-and-hold circuit is provided that includes a plurality of sample-and-hold branches arranged in parallel and each including a buffer and a sample-and-hold block including one or more sample-and-hold cells. The sample-and-hold circuit further includes a clock and timing circuit arranged for setting an adaptable time delay to enable sampling and sampling phase for each sample-and-hold block. The time delay of at least one sample-and-hold block can be set to value bigger than one sampling clock period.