Abstract: Methods and apparatus are provided for training an artificial neural network having a succession of neuron layers with interposed synaptic layers each having a respective set of N-bit fixed-point weights {w} for weighting signals propagated between its adjacent neuron layers, via an iterative cycle of signal propagation and weight-update calculation operations. Such a method includes, for each synaptic layer, storing a plurality p of the least-significant bits of each N-bit weight w in digital memory, and storing the next n-bit portion of each weight w in an analog multiply-accumulate unit comprising an array of digital memory elements. Each digital memory element comprises n binary memory cells for storing respective bits of the n-bit portion of a weight, where n?1 and (p+n+m)=N where m?0 corresponds to a defined number of most-significant zero bits in weights of the synaptic layer.

Abstract: Artificial neuron apparatus includes first and second resistive memory cells. The first resistive memory cell is connected in first circuitry having a first input and output. The second resistive memory cell is connected in second circuitry having a second input and output. The first and second circuitry are operable in alternating read and write phases to apply a programming current to their respective memory cells on receipt of excitatory and inhibitory neuron input signals, respectively. During the write phase, resistance of the respective cells is changed in response to successive excitatory and inhibitory neuron input signals. During the read phase, a read current is applied to their respective cells to produce first and second measurement signals, respectively. An output circuit connected to the first and second outputs produces a neuron output signal at a neuron output when a difference between the first and second measurement signals traverses a threshold.

Abstract: Methods and apparatus are provided for training an artificial neural network having a succession of neuron layers with interposed synaptic layers each having a respective set of N-bit fixed-point weights {w} for weighting signals propagated between its adjacent neuron layers, via an iterative cycle of signal propagation and weight-update calculation operations. Such a method includes, for each synaptic layer, storing a plurality p of the least-significant bits of each N-bit weight w in digital memory, and storing the next n-bit portion of each weight w in an analog multiply-accumulate unit comprising an array of digital memory elements. Each digital memory element comprises n binary memory cells for storing respective bits of the n-bit portion of a weight, where n?1 and (p+n+m)=N where m?0 corresponds to a defined number of most-significant zero bits in weights of the synaptic layer.

Abstract: A matrix-vector multiplication device includes a memory crossbar array with row lines, column lines, and junctions. Each junction comprises a programmable resistive element and an access element. A signal generator is configured to apply programming signals to the resistive elements to program conductance values for the matrix-vector multiplication and a readout circuit is configured to apply read voltages to the row lines and to read out current values of the column lines. Control circuitry is configured to control the signal generator and the readout circuit and to select, via the access terminals, a plurality of resistive elements in parallel according to a predefined selection scheme which applies the signals and/or the read voltages in parallel to resistive elements which do not share the same row and column line and applies the programming signals and/or the read voltages to at most one resistive element per row line and column line.

Abstract: A matrix-vector multiplication device includes a memory crossbar array with row lines, column lines, and junctions. Each junction comprises a programmable resistive element and an access element. A signal generator is configured to apply programming signals to the resistive elements to program conductance values for the matrix-vector multiplication and a readout circuit is configured to apply read voltages to the row lines and to read out current values of the column lines. Control circuitry is configured to control the signal generator and the readout circuit and to select, via the access terminals, a plurality of resistive elements in parallel according to a predefined selection scheme which applies the signals and/or the read voltages in parallel to resistive elements which do not share the same row and column line and applies the programming signals and/or the read voltages to at most one resistive element per row line and column line.

Abstract: Artificial neuron apparatus includes first and second resistive memory cells. The first resistive memory cell is connected in first circuitry having a first input and output. The second resistive memory cell is connected in second circuitry having a second input and output. The first and second circuitry are operable in alternating read and write phases to apply a programming current to their respective memory cells on receipt of excitatory and inhibitory neuron input signals, respectively. During the write phase, resistance of the respective cells is changed in response to successive excitatory and inhibitory neuron input signals. During the read phase, a read current is applied to their respective cells to produce first and second measurement signals, respectively. An output circuit connected to the first and second outputs produces a neuron output signal at a neuron output when a difference between the first and second measurement signals traverses a threshold.

Abstract: Artificial neuron apparatus includes first and second resistive memory cells. The first resistive memory cell is connected in first circuitry having a first input and output. The second resistive memory cell is connected in second circuitry having a second input and output. The first and second circuitry are operable in alternating read and write phases to apply a programming current to their respective memory cells on receipt of excitatory and inhibitory neuron input signals, respectively. During the write phase, resistance of the respective cells is changed in response to successive excitatory and inhibitory neuron input signals. During the read phase, a read current is applied to their respective cells to produce first and second measurement signals, respectively. An output circuit connected to the first and second outputs produces a neuron output signal at a neuron output when a difference between the first and second measurement signals traverses a threshold.

Abstract: Artificial neuron apparatus includes a resistive memory cell connected in an input circuit having a neuron input, for receiving neuron input signals, and a current source for supplying a read current to the cell. The input circuit is selectively configurable in response to a set of control signals, defining alternating read and write phases of operation, to apply the read current to the cell during the read phase and to apply a programming current to the cell, for programming cell resistance, on receipt of a neuron input signal during the write phase. The cell resistance is progressively changed from a first state to a second state in response to successive neuron input signals. The apparatus further includes an output circuit comprising a neuron output and a digital latch which is connected to the input circuit for receiving a measurement signal dependent on cell resistance.

Abstract: Artificial neuron apparatus includes first and second resistive memory cells. The first resistive memory cell is connected in first circuitry having a first input and output. The second resistive memory cell is connected in second circuitry having a second input and output. The first and second circuitry are operable in alternating read and write phases to apply a programming current to their respective memory cells on receipt of excitatory and inhibitory neuron input signals, respectively. During the write phase, resistance of the respective cells is changed in response to successive excitatory and inhibitory neuron input signals. During the read phase, a read current is applied to their respective cells to produce first and second measurement signals, respectively. An output circuit connected to the first and second outputs produces a neuron output signal at a neuron output when a difference between the first and second measurement signals traverses a threshold.

Abstract: Artificial neuron apparatus includes a resistive memory cell connected in an input circuit having a neuron input, for receiving neuron input signals, and a current source for supplying a read current to the cell. The input circuit is selectively configurable in response to a set of control signals, defining alternating read and write phases of operation, to apply the read current to the cell during the read phase and to apply a programming current to the cell, for programming cell resistance, on receipt of a neuron input signal during the write phase. The cell resistance is progressively changed from a first state to a second state in response to successive neuron input signals. The apparatus further includes an output circuit comprising a neuron output and a digital latch which is connected to the input circuit for receiving a measurement signal dependent on cell resistance.

Abstract: A novel and useful method of visualization by detection of EM radiation being irradiated or reflected from objects in the imager's field of view using Finite Element Method (FEM) simulation software tools. The methodology provides a verification method of antenna operation from an electrical point of view since bolometer performance cannot be estimated using regular antenna parameters such as directivity, gain, impedance matching, etc. as the bolometer does not behave as an antenna but rather behaves as an absorber. An incident wave is triggered on the absorber and the absorption of the bolometer structure is estimated using commercially available Finite Element Method (FEM) software (e.g., ANSYS® HFSS software, CST MICROWAVE STUDIO®, etc.). How much of the energy is reflected is subsequently measured. The energy which is not reflected is considered to be absorbed by the absorber.

Abstract: The exemplary embodiments relate to an asynchronously clocked successive approximation register analog-to-digital converter (SAR ADC) configured to provide a digital approximation of a sampled input signal as a result of an asynchronous successive approximation operation. The converter includes a regulation circuit configured to determine whether the asynchronous successive approximation operation was performed within a predefined conversion time and to regulate the SAR ADC such that the conversion time of the asynchronous operation is shifted towards the predefined conversion time. The embodiments further relate to a corresponding method and a corresponding design structure.

Abstract: A zero-crossing amplifier unit for use in high speed analog-digital-converters. A gain stage compares a sampling voltage at an input node with a provided threshold voltage to obtain a gain stage output signal. A voltage controlled current source provides a load current depending on a time window between an initial slope and an end slope of the gain stage output signal. A slope control means increases a duration of a rise and/or fall time of at least one of the initial and end slopes of the gain stage output signal.

Abstract: Methods, computing systems and computer program products implement embodiments of the present invention that include detecting multiple sets of storage objects stored in a data facility including multiple server racks, each of the server racks including a plurality of server computers, each of the storage objects in each set being stored in a separate one of the server racks and including one or more data objects and one or more protection objects. A specified number of the storage objects are identified in a given server rack, each of the identified storage objects being stored in a separate one of the server computers, and one or more server computers in the given server rack not storing any of the identified storage objects are identified. Finally, in the identified one or more server computers, an additional protection object is created and managed for the identified storage objects.

Abstract: A charge sharing circuit for generating a calibration voltage. The circuit comprises a calibration capacitor for providing at an upper terminal of the calibration capacitor the calibration voltage. The circuit further comprises a series connection of a plurality of N switches, wherein N is an integer>2, and a plurality of at least N?1 switching capacitors. Each switching capacitor is coupled to one connecting node connecting two of the N switches. One side of the series connection of the plurality of N switches is coupled to the upper terminal of the calibration capacitor and the other side of the series connection of the N switches is coupled to a fixed voltage. The circuit is configured to transmit at least two clock signals to selectively drive at least two distinct subsets of the switches. There is further provided a corresponding method and a corresponding design structure.

Abstract: The present invention relates to a successive approximation register analog-to-digital converter (SAR ADC) for providing a digital approximation of a sampled differential input signal as a result of a successive approximation operation. The SAR ADC comprises a first comparison stage configured to perform a first set of decision steps of the successive approximation operation and a second comparison stage configured to perform a second set of decision steps of the successive approximation operation. Furthermore, the SAR ADC comprises a regulation circuit configured to adjust the common mode of the input signal towards a target common mode before the second comparison stage performs the second set of decision steps. The present invention further relates to a corresponding method and a corresponding design structure.

Abstract: A linear reference analog to digital converter (ADC) network may include a first ADC operatively connected to a first sample and hold circuit. The linear reference ADC network may be configured to receive an input signal from the first sample and hold circuit and sample the input signal with a harmonic distortion. The linear reference ADC network may further include a reference ADC operatively connected to a second sample and hold circuit and configured to receive the input signal and sample the input signal with a second harmonic distortion. The linear reference ADC network may further include a combining module operatively connected to the first ADC and the reference ADC, the combining module configured to equalize a linearity of an output of the first ADC to a linearity of an output of the reference ADC, and output a combined output signal, and a circuit configured to output a calibrated output signal having calibrated harmonic distortion content.

Abstract: A comparator for an analog-to-digital converter is provided. The comparator includes a differential amplifier unit that receives a sampling signal and provides an output signal, based on a voltage provided by the sampling signal. The differential amplifier unit includes an input stage that receives the sampling signal and integrates a current on the integration nodes based on potentials of the sampling signal. The comparator includes a sense amplifier coupled with the integration nodes that detects a potential difference and amplifies the potential difference to generate the output signal. The comparator includes a charge injection circuit (30) to inject equal charges into the integration nodes.

Abstract: A distributed electrostatic discharge protection circuit includes a plurality of electrostatic discharge protection elements and a current balancing network connecting the plurality of electrostatic discharge protection elements. The current balancing network is configured in a return path of the distributed electrostatic discharge protection circuit such that during an electrostatic discharge (ESD) event, the circuit provides predefined current density within each of the electrostatic discharge protection elements.

Abstract: An analog-digital converter with successive approximation includes a capacitor array for being loaded by applying a given input signal potential and for providing a sampling potential, wherein capacitors of the capacitor array are serially coupled with switches. A decision latch is included for evaluating the sampling potential in a number of consecutive decision steps. The analog-digital converter also includes a logic unit for selectively changing the sampling potential by selectively switching switches associated to the capacitors of the capacitor array for each decision step based on an evaluation result of a previous decision step, wherein the switches are respectively coupled with a calibration switch.