A neuromorphic computing device includes first and second memory cell arrays, and an analog-to-digital converting circuit. The first memory cell array includes a plurality of resistive memory cells, generates a plurality of read currents based on a plurality of input signals and a plurality of data, and outputs the plurality of read currents through a plurality of bitlines or source lines. The second memory cell array includes a plurality of reference resistive memory cells and an offset resistor, and outputs a reference current through a reference bitline or a reference source line. The analog-to-digital converting circuit converts the plurality of read currents into a plurality of digital signals based on the reference current. The offset resistor is connected between the reference bitline and the reference source line.

Certain aspects provide a circuit for in-memory computation. The circuit generally includes an in-memory computation array having a plurality of computation circuits, each of the computation circuits being configured to perform a dot product computation. In certain aspects, each of the computation circuits includes a memory cell, a capacitive element, a precharge transistor coupled between an output of the memory cell and the capacitive element, and a read transistor coupled between a read bit line (RBL) and the capacitive element.

A device configured to implement an artificial intelligence deep neural network includes a first matrix and a second matrix. The first matrix resistive processing unit (“RPU”) array receives a first input vector along the rows of the first matrix RPU. A second matrix RPU array receives a second input vector along the rows of the second matrix RPU. A reference matrix RPU array receives an inverse of the first input vector along the rows of the reference matrix RPU and an inverse of the second input vector along the rows of the reference matrix RPU. A plurality of analog to digital converters are coupled to respective outputs of a plurality of summing junctions that receive respective column outputs of the first matrix RPU array, the second matrix RPU array, and the reference RPU array and provides a digital value of the output of the plurality of summing junctions.

A device configured to implement an artificial intelligence deep neural network includes a first matrix and a second matrix. The first matrix resistive processing unit (“RPU”) array receives a first input vector along the rows of the first matrix RPU. A second matrix RPU array receives a second input vector along the rows of the second matrix RPU. A reference matrix RPU array receives an inverse of the first input vector along the rows of the reference matrix RPU and an inverse of the second input vector along the rows of the reference matrix RPU. A plurality of analog to digital converters are coupled to respective outputs of a plurality of summing junctions that receive respective column outputs of the first matrix RPU array, the second matrix RPU array, and the reference RPU array and provides a digital value of the output of the plurality of summing junctions.

A memory circuit configured to perform multiply-accumulate (MAC) operations for performance of an artificial neural network includes a series of synapse cells arranged in a cross-bar array. Each cell includes a memory transistor connected in series with a memristor. The memory circuit also includes input lines connected to the source terminal of the memory transistor in each cell, output lines connected to an output terminal of the memristor in each cell, and programming lines coupled to a gate terminal of the memory transistor in each cell. The memristor of each cell is configured to store a conductance value representative of a synaptic weight of a synapse connected to a neuron in the artificial neural network, and the memory transistor of each cell is configured to store a threshold voltage representative of a synaptic importance value of the synapse connected to the neuron in the artificial neural network.

A memory circuit configured to perform multiply-accumulate (MAC) operations for performance of an artificial neural network includes a series of synapse cells arranged in a cross-bar array. Each cell includes a memory transistor connected in series with a memristor. The memory circuit also includes input lines connected to the source terminal of the memory transistor in each cell, output lines connected to an output terminal of the memristor in each cell, and programming lines coupled to a gate terminal of the memory transistor in each cell. The memristor of each cell is configured to store a conductance value representative of a synaptic weight of a synapse connected to a neuron in the artificial neural network, and the memory transistor of each cell is configured to store a threshold voltage representative of a synaptic importance value of the synapse connected to the neuron in the artificial neural network.

A data processing system may include: a controller configured to receive a neural network processing request from a host device; a processing memory including: one or more sub arrays each including memory cells coupled between row lines and column lines; multiplexers (MUXs) provided for respective column line groups, which are configured by grouping the column lines by a preset number; and analog-to-digital converters (ADCs) coupled to the respective MUXs; and a deserializer. The deserializer is configured to receive, from the controller, data to be stored in a selected sub array and a first column address at which the data is to be stored, and remap the first column address to a second column address such that the data is distributed and stored in the memory cells coupled to the column line groups, in order to store the data in the sub array.

A data processing system may include: a controller configured to receive a neural network processing request from a host device; a processing memory including: one or more sub arrays each including memory cells coupled between row lines and column lines; multiplexers (MUXs) provided for respective column line groups, which are configured by grouping the column lines by a preset number; and analog-to-digital converters (ADCs) coupled to the respective MUXs; and a deserializer. The deserializer is configured to receive, from the controller, data to be stored in a selected sub array and a first column address at which the data is to be stored, and remap the first column address to a second column address such that the data is distributed and stored in the memory cells coupled to the column line groups, in order to store the data in the sub array.

A memory circuit for a print component including a plurality of I/O pads, including an analog pad, to connect to a plurality of signal paths which communicate operating signals to the print component. The memory circuit includes a controllable selector connected in line with one of the signal paths via the I/O pads, the selector controllable to disconnect the corresponding signal path to the print component, and a memory component to store memory values associated with the print component. A control circuit, in response to a sequence of operating signals received by the I/O pads representing a memory read, to operate the controllable selector to disconnect the signal path to the print component to block the memory read of the print component, and provide an analog signal to the analog pad to provide an analog electrical value at the analog pad representing stored memory values selected by the memory read.

Numerous examples of an input function circuit block and an output neuron circuit block coupled to a vector-by-matrix multiplication (VMM) array in an artificial neural network are disclosed. In one example, an artificial neural network comprises a vector-by-matrix multiplication array comprising a plurality of non-volatile memory cells organized into rows and columns; an input function circuit block to receive digital input signals, convert the digital input signals into analog signals, and apply the analog signals to control gate terminals of non-volatile memory cells in one or more rows of the array during a programming operation; and an output neuron circuit block to receive analog currents from the columns of the array during a read operation and generate an output signal.