Systems are provided for implementing a hardware accelerator. The hardware accelerator emulate a stochastic neural network, and includes a first memristor crossbar array, and a second memristor crossbar array. The first memristor crossbar array can be programmed to calculate node values of the neural network. The nodes values can be calculated in accordance with rules to reduce an energy function associated with the neural network. The second memristor crossbar array is coupled to the first memristor crossbar array and programmed to introduce noise signals into the neural network. The noise signals can be introduced such that the energy function associated with the neural network converges towards a global minimum and modifies the calculated node values.

In one example, a device to process analog sensor data is described. For example, a device may include at least one analog sensor to generate a first set of analog voltage signals and a crossbar array including a plurality of memristors. In one example, the crossbar array is to receive an input vector of the first set of analog voltage signals, generate an output vector comprising a second set of analog voltage signals that is based upon a dot product of the input vector and a matrix comprising resistance values of the plurality of memristors, detect a pattern of the output vector, and activate a processor upon a detection of the pattern.

A semiconductor device is provided. The semiconductor device includes a substrate a substrate, a first electrode structure on the substrate, the first electrode structure including first insulating patterns and first electrode patterns, the first insulating patterns alternately stacked with the first electrode patterns, a second electrode pattern on a sidewall of the first electrode structure, and a data storage film on a sidewall of the second electrode pattern. The data storage film has a variable resistance.

A resistive random access memory includes a memory cell including a resistive element having a resistance which varies according to a write operation and stores data according to the resistance of the resistive element, a reference resistive element having a resistance set to a first value, a voltage line set to a first voltage during a first write operation in which the resistance of the resistive element is varied from a second value higher than the first value to the first value, and a voltage control circuit arranged between first ends of the two resistive elements. The voltage control circuit adjusts a value of the first voltage supplied from the voltage line so as to reduce a difference between currents flowing through the two resistive elements during the first write operation, and supply the adjusted first voltage to the first ends of the two resistive elements.

A system includes a cross-point memory array and a decoder circuit coupled to the cross-point memory array. The decoder circuit includes a predecoder having predecode logic to generate a control signal and a level shifter circuit to generate a voltage signal. The decoder circuit further includes a post-decoder coupled to the predecoder, the post-decoder including a first stage and a second stage coupled to the first stage, the control signal to control the first stage and the second stage to route the voltage signal through the first stage and the second stage to a selected conductive array line of a plurality of conductive array lines coupled to a memory array.

Methods, systems, and devices for self-selecting memory with horizontal access lines are described. A memory array may include first and second access lines extending in different directions. For example, a first access line may extend in a first direction, and a second access line may extend in a second direction. At each intersection, a plurality of memory cells may exist, and each plurality of memory cells may be in contact with a self-selecting material. Further, a dielectric material may be positioned between a first plurality of memory cells and a second plurality of memory cells in at least one direction. each cell group (e.g., a first and second plurality of memory cells) may be in contact with one of the first access lines and second access lines, respectively.

Methods and apparatuses for a cross-point memory array and related fabrication techniques are described. The fabrication techniques described herein may facilitate concurrently building two or more decks of memory cells disposed in a cross-point architecture. Each deck of memory cells may include a plurality of first access lines (e.g., word lines), a plurality of second access lines (e.g., bit lines), and a memory component at each topological intersection of a first access line and a second access line. The fabrication technique may use a pattern of vias formed at a top layer of a composite stack, which may facilitate building a 3D memory array within the composite stack while using a reduced number of processing steps. The fabrication techniques may also be suitable for forming a socket region where the 3D memory array may be coupled with other components of a memory device.

A device may include a first die having a first circuit and a second die having a second circuit. The die may be separated by a material layer. The material layer may include multiple through-silicon vias (TSVs) for electrically coupling the first die to the second die. A first TSV of the TSVs may electrically couple the first circuit to the second circuit and a second TSV of the TSVs may include a redundant TSV that electrically bypasses the first TSV to couple the first circuit to the second circuit if a fault is detected in the first TSV.

Aspects of the present disclosure provide a method for calibrating crossbar-based apparatuses. The method includes obtaining output data of a crossbar-based apparatus may include a plurality of cross-point devices with tunable conductance, where the output data of the crossbar-based apparatus represents computing results of at least one operation performed by the crossbar-based apparatus, and where the output data corresponding to a plurality of settings of a plurality of analog components of the crossbar-based apparatus. The method also includes obtaining, by a processing device, one or more calibration parameters based on the output data of the crossbar-based apparatus, where the one or more calibration parameters correspond to one or more errors associated with one or more of the analog components of the crossbar-based apparatus. The method further includes calibrating the crossbar-based apparatus using the one or more calibration parameters.

A circuit for performing energy-efficient and high-throughput multiply-accumulate (MAC) arithmetic dot-product operations and convolution computations includes a two dimensional crossbar array comprising a plurality of row inputs and at least one column having a plurality of column circuits, wherein each column circuit is coupled to a respective row input. Each respective column circuit includes an excitatory memristor neuron circuit having an input coupled to a respective row input, a first synapse circuit coupled to an output of the excitatory memristor neuron circuit, the first synapse circuit having a first output, an inhibitory memristor neuron circuit having an input coupled to the respective row input, and a second synapse circuit coupled to an output of the inhibitory memristor neuron circuit, the second synapse circuit having a second output. An output memristor neuron circuit is coupled to the first output and second output of each column circuit and has an output.