Abstract: A computing system having a computational memory and a method configured to perform computations using an approximate message passing process. The system exploits memcomputing which is a prominent non-von Neumann computational approach expected to significantly improve an energy efficiency of computing systems. The computational memory includes at least one memristive array comprising a plurality of memristive devices arranged in a crossbar topology and the computing system may further comprise digital combinational control circuitry adapted to perform read and write operations on the at least one memristive array and to store at least one state variable of the approximate message passing process. An output of the at least one memristive array represents a result of a computation of the approximate message passing process. The control circuitry may comprise circuitry to iteratively perform computations that may not require high precision.

Abstract: The invention is notably directed to a resistive memory device comprising a control unit for controlling the resistive memory device and a plurality of memory cells. The plurality of memory cells includes a first terminal, a second terminal and a phase change segment comprising a phase-change material for storing information in a plurality of resistance states. The phase change segment is arranged between the first terminal and the second terminal. The phase change material consists of antimony. Furthermore, at least one of the dimensions of the phase change segment is smaller than 15 nanometers. Additional implementations of the resistive memory device include a related method, a related control unit, a related memory cell and a related computer program product.

Abstract: Neuromorphic synapse apparatus is provided comprising a synaptic device and a control signal generator. The synaptic device comprises a memory element, disposed between first and second terminals, for conducting a signal between those terminals with an efficacy which corresponds to a synaptic weight in a read mode of operation, and a third terminal operatively coupled to the memory element. The memory element has a non-volatile characteristic, which is programmable to vary the efficacy in response to programming signals applied via the first and second terminals in a write mode of operation, and a volatile characteristic which is controllable to vary the efficacy in response to control signals applied to the third terminal. The control signal generator is responsive to input signals and is adapted to apply control signals to the third terminal in the read and write modes, in dependence on the input signals, to implement predetermined synaptic dynamics.

Abstract: A neuromorphic processing device has a device input, for receiving an input data signal, and an assemblage of neuron circuits. Each neuron circuit comprises a resistive memory cell which is arranged to store a neuron state, indicated by cell resistance, and to receive neuron input signals for programming cell resistance to vary the neuron state, and a neuron output circuit for supplying a neuron output signal in response to cell resistance traversing a threshold. The device includes an input signal generator, connected to the device input and the assemblage of neuron circuits, for generating neuron input signals for the assemblage in dependence on the input data signal. The device further includes a device output circuit, connected to neuron output circuits of the assemblage, for producing a device output signal dependent on neuron output signals of the assemblage, whereby the processing device exploits stochasticity of resistive memory cells of the assemblage.

Abstract: A multiplication device for performing a matrix-vector-multiplication may be provided. The multiplication device comprises a memristive crossbar array comprising a plurality of memristive devices. The device comprises a decomposition unit adapted for decomposing a matrix into a partial sum of multiple sub-matrices, and decomposing a vector into a sum of multiple sub-vectors, a programming unit adapted for programming the plurality of the memristive devices with values representing elements of the sub-matrices such that each one of the memristive devices corresponds to one of the elements of the sub-matrices, an applying unit adapted for applying elements of one of the multiple sub-vectors as input values to the memristive crossbar array to input lines of the memristive crossbar array resulting in partial results at output lines of the memristive crossbar array, and a summing unit adapted for scaling and summing the partial results building the product of the matrix and the vector.

Abstract: A multiplication device for performing a matrix-vector-multiplication may be provided. The multiplication device comprises a memristive crossbar array comprising a plurality of memristive devices. The device comprises a decomposition unit adapted for decomposing a matrix into a partial sum of multiple sub-matrices, and decomposing a vector into a sum of multiple sub-vectors, a programming unit adapted for programming the plurality of the memristive devices with values representing elements of the sub-matrices such that each one of the memristive devices corresponds to one of the elements of the sub-matrices, an applying unit adapted for applying elements of one of the multiple sub-vectors as input values to the memristive crossbar array to input lines of the memristive crossbar array resulting in partial results at output lines of the memristive crossbar array, and a summing unit adapted for scaling and summing the partial results building the product of the matrix and the vector.

Abstract: A circuit implementing a spiking neural network that includes a learning component that can learn from temporal correlations in the spikes regardless of correlations in the rates. In some embodiments, the learning component comprises a rate-discounting component. In some embodiments, the learning rule computes a rate-normalized covariance (normcov) matrix, detects clusters in this matrix, and sets the synaptic weights according to these clusters. In some embodiments, a synapse with a long-term plasticity rule has an efficacy that is composed by a weight and a fatiguing component. In some embodiments, A Hebbian plasticity component modifies the weight component and a short-term fatigue plasticity component modifies the fatiguing component. The fatigue component increases with increases in the presynaptic spike rate. In some embodiments, the fatigue component increases are implemented in a spike-based manner.

Abstract: A circuit implementing a spiking neural network that includes a learning component that can learn from temporal correlations in the spikes regardless of correlations in the rates. In some embodiments, the learning component comprises a rate-discounting component. In some embodiments, the learning rule computes a rate-normalized covariance (normcov) matrix, detects clusters in this matrix, and sets the synaptic weights according to these clusters. In some embodiments, a synapse with a long-term plasticity rule has an efficacy that is composed by a weight and a fatiguing component. In some embodiments, A Hebbian plasticity component modifies the weight component and a short-term fatigue plasticity component modifies the fatiguing component. The fatigue component increases with increases in the presynaptic spike rate. In some embodiments, the fatigue component increases are implemented in a spike-based manner.

Abstract: A computing system includes computational memory and digital combinational circuitry operatively coupled with the computational memory. The computational memory is configured to perform computations at a prescribed precision. The digital combinational circuitry is configured to increase the precision of the computations performed by the computational memory. The computational memory and the digital combinational circuitry may be configured to iteratively perform a computation to a predefined precision. The computational memory may include circuitry configured to perform analog computation using values stored in the computational memory, and the digital combinational circuitry may include a central processing unit, a graphics processing unit and/or application specific circuitry. The computational memory may include an array of resistive memory elements having resistance or conductance values stored therein, the respective resistance or conductance values being programmable.

Abstract: The invention is notably directed to a device for performing a matrix-vector multiplication of a matrix with a vector. The device comprises a memory crossbar array comprising of row lines, of columns lines and of junctions arranged between the row lines and the column lines. Each junction comprises a programmable resistive memory element. The device comprises a signal generator and a readout circuit. The device is configured to perform a calibration procedure to compensate for conductance variations of the resistive memory elements. The calibration procedure is configured to program a calibration subset of the plurality of resistive memory elements to initial conductance values and to apply a constant calibration voltage to the row lines of the calibration subset. The device is configured to read calibration current values of the column lines of the calibration subset and to derive an estimation of a conductance variation parameter from the calibration current values.

Abstract: Apparatus including: memory cell unit(s) having a variable-resistance channel component (CC) extending between first and second supply terminals for supplying read and write (R/W) signals to the unit in respective R/W modes, and resistive memory elements (RMEs) arranged along the CC, RME includes resistive memory material (RMM), extending along a respective channel segment (CHS) of the CC in contact therewith, in which respective lengths along that CHS of high- and low-resistance regions is variable in write mode, and a gate terminal provided on that CHS for controlling resistance of the CHS in response to control signal(s) (CS) applied to the gate terminal; and circuitry configured to apply the CS such that, in read mode, a RME(s) is selected by applying a CS producing CHS with resistance between the resistance regions of the RMM; and remaining RME(s) are deselected by applying CS producing CHS having resistance less than the low-resistance region.

Abstract: Apparatus including: memory cell unit(s) having a variable-resistance channel component (CC) extending between first and second supply terminals for supplying read and write (R/W) signals to the unit in respective R/W modes, and resistive memory elements (RMEs) arranged along the CC, RME includes resistive memory material (RMM), extending along a respective channel segment (CHS) of the CC in contact therewith, in which respective lengths along that CHS of high- and low-resistance regions is variable in write mode, and a gate terminal provided on that CHS for controlling resistance of the CHS in response to control signal(s) (CS) applied to the gate terminal; and circuitry configured to apply the CS such that, in read mode, a RME(s) is selected by applying a CS producing CHS with resistance between the resistance regions of the RMM; and remaining RME(s) are deselected by applying CS producing CHS having resistance less than the low-resistance region.

Abstract: An apparatus for programming at least one multi-level Phase Change Memory (PCM) cell having a first terminal and a second terminal. A programmable control device controls the PCM cell to have a respective cell state by applying at least one current pulse to the PCM cell, the control device controlling the at least one current pulse by applying a respective first pulse to the first terminal and a respective second pulse applied to the second terminal of the PCM cell. The respective cell state is defined by a respective resistance level. The control device receives a reference resistance value defining a target resistance level for the cell, and further receives an actual resistance value of said PCM cell such that the applying the respective first pulse and said respective second pulse is based on said actual resistance value of the PCM cell and said received reference resistance value.

Abstract: Method to produce a neuromorphic synapse apparatus comprising a memelement for storing a synaptic weight, and programming logic. The memelement is adapted to exhibit a desired programming characteristic. The programming logic is responsive to a stimulus prompting update of the synaptic weight for generating a programming signal for programming the memelement to update said weight. The programming logic may be responsive to an input signal indicating an input weight-change value ?Wi, and may be adapted to generate a programming signal dependent on the input weight-change value ?Wi. The programming logic is adapted such that the programming signals exploit the programming characteristic of the memelement to provide a desired weight-dependent synaptic update efficacy.

Abstract: Neuromorphic synapse apparatus 11 comprises a memelement 20 for storing a synaptic weight, and programming logic 21. The memelement 20 is adapted to exhibit a desired programming characteristic. The programming logic 21 is responsive to a stimulus prompting update of the synaptic weight for generating a programming signal for programming the memelement 20 to update said weight. The programming logic 21 may be responsive to an input signal indicating an input weight-change value ?Wi, and may be adapted to generate a programming signal dependent on the input weight-change value ?Wi. The programming logic 21 is adapted such that the programming signals exploit the programming characteristic of the memelement 20 to provide a desired weight-dependent synaptic update efficacy.

Abstract: A sensor arrangement for position sensing comprises a magnetic field source and a magnetoresistive element arranged in a magnetic field generated by the magnetic field source, which magnetoresistive element provides an output signal (R) dependent on a position (x) of the magnetoresistive element relative to the magnetic field source. A feedback controller is configured to receive the output signal (R) of the magnetoresistive element and is configured to adjust one or more of the position (x) of the magnetoresistive element relative to the magnetic field source and a strength of the magnetic field generated by the magnetic field source acting on the magnetoresistive element dependent on the output signal (R) of the magnetoresistive element.

Abstract: A sensor arrangement for position sensing comprises a row of multiple magnetoresistive elements. A magnetic field source (3) provides a magnetic field with a first magnetic pole (N) and a second magnetic pole (S). The magnetic field source (3) is arranged such that magnetoresistive elements of the row face one of: the first magnetic pole (N) or second magnetic pole (S). The first magnetoresistive element is arranged in the magnetic field and provides a first output signal dependent on a position of the magnetoresistive element relative to the magnetic field source (3). A measurement unit is configured to determine a position of the magnetic field source (3) relative to the magnetoresistive elements of the row dependent on the first output signals of the magnetoresistive elements.

Abstract: A device for determining an actual cell state of a resistive memory cell having a plurality M of programmable cell states comprising a sensing circuit, a settling circuit, a prebiasing circuit, and a resistor coupled in parallel to the resistive memory cell, wherein the resistor is configured to reduce an effective resistance seen by the prebiasing circuit. The sensing circuit is configured to sense a sensing voltage of the resistive memory cell and output a resultant value in response to the sensing voltage which is indicative for the actual cell state. The settling circuit is configured to settle the sensing voltage to a certain target voltage representing one of the M programmable cell states. The prebiasing circuit is configured to prebiase a bitline capacitance of the resistive memory cell such the sensing voltage is close to the certain target voltage.

Abstract: A method for determining of a position of an object using a sensor arrangement that includes a first magnetoresistive element and a second magnetoresistive element. A source provides a magnetic field with first and second magnetic poles. The source is arranged between the first magnetoresistive element and the second magnetoresistive element with the first magnetic pole facing the first magnetoresistive element and the second magnetic pole facing the second magnetoresistive element. The first magnetoresistive element is arranged in the magnetic field and provides a first output signal dependent on a position of the first magnetoresistive element relative to the magnetic field source. The second magnetoresistive element is arranged in the magnetic field and provides a second output signal dependent on a position of the second magnetoresistive element relative to the magnetic field source.

Abstract: A method of fabricating a resistive memory element having a layer structure includes: providing a substrate; depositing a first electrode on an upper surface of the substrate; forming a layer of confining material on an upper surface of the first electrode so as to define a cavity having a maximal lateral dimension that is less than 60 nm along a direction parallel to an average plane of the first electrode, the confining material having a thermal conductivity greater than 0.5 W/(m·K); depositing a resistively switchable material as an amorphous compound comprising carbon to fill the cavity; and depositing a second electrode on an upper surface of the resistively switchable material.