# US 2020/0401373 A1 · Variable accuracy computing system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2020/0401373 A1
- Document ID: US20200401373A1
- Status: Published application
- Kind: A1
- Application number: 15/931,860
- Filed: 2020-05-14
- Published: 2020-12-24
- Earliest claimed: 2019-06-18
- Assignees: Cirrus Logic International Semiconductor Ltd.
- Applicants: Cirrus Logic International Semiconductor Ltd.
- Inventors: John Paul LESSO
- Web page: https://patentgrep.com/patent/US20200401373A1

## CPC

- G06F 7/523: Multiplying only
- G06F 7/5443: {Sum of products (for applications thereof, see the relevant places, e.g. G06F17/10, H03H17/00)}
- G06F 1/3203: Power management, i.e. event-based initiation of a power-saving mode
- G06F 1/3243: {Power saving in microcontroller unit}
- G06G 7/16: for multiplication or division {(G06G7/19 and G06G7/24 take precedence; measuring electric power G01R21/00)}
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/0499: Feedforward networks
- G06N 3/065: Analogue means
- G06F 11/3024: {where the computing system component is a central processing unit [CPU]}
- G06F 11/3058: {Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations (thermal management in cooling arrangements of a computing system G06F1/206)}
- G06F 2207/382: Reconfigurable for different fixed word lengths
- G06F 2207/4814: Non-logic devices, e.g. operational amplifiers
- G06F 2207/4824: Neural networks
- G06N 3/045: Combinations of networks
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06G: ANALOGUE COMPUTERS (analogue optical computing devices G06E3/00)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

The present disclosure relates to a computing system. The computing system comprises a data input configured to receive an input data signal, a computation unit having an input coupled with the data input, the computation unit being operative to apply a weight to a signal received at its input to generate a weighted output signal, and a controller. The controller is configured to monitor a parameter of the input signal and/or a parameter of the output signal and to issue a control signal to the computation unit to control a level of accuracy of the weighted output signal based at least in part on the monitored parameter.

## Claims

### Claim 1

A computing system comprising:
a data input configured to receive an input data signal;
a computation unit having an input coupled with the data input, the computation unit being operative to apply a weight to a signal received at its input to generate a weighted output signal; and
a controller,
wherein the controller is configured to monitor a parameter of the input signal and/or a parameter of the output signal and to issue a control signal to the computation unit to control a level of accuracy of the weighted output signal based at least in part on the monitored parameter.

### Claim 2 (depends on 1)

A computing system according to claim 1 wherein in response to the control signal the computation unit is operative to adjust a number of bits of the input data signal that are used by the computation unit to generate the weighted output signal.

### Claim 3 (depends on 2)

A computing system according to claim 2 wherein the computation unit is operative to prevent use of one or more least significant bits of the input data signal from being used by the computation unit to generate the weighted output signal.

### Claim 4 (depends on 1)

A computing system according to claim 1 wherein in response to the control signal the computation unit is operative to adjust a number of bits of the weight that are used by the computation unit to generate the weighted output signal.

### Claim 5 (depends on 4)

A computing system according to claim 4 wherein the computation unit is operative to prevent use of one or more least significant bits of the weight from being used by the computation unit to generate the weighted output signal.

### Claim 6 (depends on 1)

A computing system according to claim 1 wherein in response to the control signal the computation unit is operative to adjust a number of bits that are provided for the weighted output signal.

### Claim 7 (depends on 1)

A computing system according to claim 1 wherein the input data signal comprises an audio signal.

### Claim 8 (depends on 1)

A computing system according to claim 1 wherein the parameter of the input signal monitored by the controller comprises one or more of:
a signal to noise ratio;
a noise level;
a direct to reverberate ratio;
an average spectrum;
a spectral tilt;
Mel-frequency cepstral coefficients (MFCCs) or a property derived therefrom;
a long-term spectrum of the input signal;
an indication of narrowband interference in the input signal;
an indication of voice activity detection (VAD) in the input signal; and/or
an indication of automatic speech recognition (ASR) confidence for the input signal.

### Claim 9 (depends on 1)

A computing system according to claim 1 wherein the parameter of the output signal monitored by the controller comprises one or more of:
a metric indicative of class discriminability of the output of the computation unit;
a difference between the output signal and one or more thresholds;
Kullback Leibler (KL) divergence;
a statistical distance metric between outputs of the computation unit;
a mean, standard deviation (STD), skew or kurtosis of the output signal in comparison to typical values.

### Claim 10 (depends on 1)

A computing system according to claim 1 wherein the computation unit comprises a crossbar array, wherein the crossbar array is configured to perform parallel computation on bits of the input data signal, or to perform computation on bits of the input data signal sequentially.

### Claim 11 (depends on 10)

A computing system according to claim 10 wherein the crossbar array comprises:
a plurality of row electrodes, each of the plurality of row electrodes having a respective input terminal;
a plurality of column electrodes, each of the plurality of column electrodes having a respective output terminal; and
a plurality of computation elements,
wherein each of the plurality of computation elements is coupled to one of the plurality of row electrodes and to one of the plurality of column electrodes, and is configured to output a signal representative of the product of a value of a signal received at the input terminal of the row electrode to which it is coupled and a weight value associated with the computation element,
and wherein in response to receiving a control signal from the controller to reduce the level of accuracy of the output of the computation unit the computation unit is operative to:
disable one or more of the row electrodes; or
disable one or more of the column electrodes.

### Claim 12 (depends on 11)

A computing system according to claim 11 wherein the crossbar array further comprises:
a plurality of multipliers, each of the plurality of multipliers having an input coupled to the output terminal of a respective one of the plurality of column electrodes; and
a summation unit coupled to outputs of the plurality of multipliers, the summation unit being operative to output a signal representing a sum of values represented by the outputs of the plurality of multipliers,
wherein each of the plurality of multipliers is configured to apply a binary weighting to a signal received at its input and to output a binary weighted signal to the summation unit.

### Claim 13 (depends on 10)

A computing system according to claim 10 wherein the crossbar array comprises:
a row electrode having an input terminal;
a plurality of column electrodes, each of the plurality of column electrodes having a respective output terminal; and
a plurality of computation elements,
wherein each of the plurality of computation elements is coupled to the row electrode and to one of the plurality of column electrodes, and is configured to output a signal representative of the product of a value of a signal received at the input terminal of the row electrode and a weight value associated with the computation element,
and wherein in response to receiving a control signal from the controller to reduce the level of accuracy of the output of the computation unit the computation unit is operative to:
disable the row electrode for a predetermined period of time; or
disable one or more of the column electrodes for a predetermined period of time.

### Claim 14 (depends on 13)

A computing system according to claim 13 wherein the crossbar array further comprises:
a plurality of multipliers, each of the plurality of multipliers having an input coupled to the output terminal of a respective one of the plurality of column electrodes; and
a summation unit coupled to outputs of the plurality of multipliers, the summation unit being operative to output a signal representing a sum of values represented by the outputs of the plurality of multipliers; and
an accumulator coupled to an output of the summation unit;
wherein each of the plurality of multipliers is configured to apply a binary weighting to a signal received at its input and to output a binary weighted signal to the summation unit.

### Claim 15 (depends on 1)

A computing system according to claim 1 wherein the computing system comprises analog neural network circuitry.

### Claim 16 (depends on 1)

A computing system according to claim 1 wherein the controller is operative to monitor an environmental parameter of the computation unit and to issue a control signal to the computation unit to control a level of accuracy of the weighted output signal based at least in part on the monitored environmental parameter, wherein the environmental parameter comprises a supply voltage or a temperature of the computation unit.

### Claim 17 (depends on 1)

A computing system according to claim 1 wherein the computation unit is a digital computation unit configured to perform bit-series computation of bits of the input data signal and the weight.

### Claim 18 (depends on 1)

An integrated circuit comprising a computing system according to claim 1.

### Claim 19 (depends on 18)

A device comprising an integrated circuit according to claim 18, wherein the device is a mobile telephone, a tablet or laptop computer or an Internet of Things (IoT) device.

### Claim 20

A computing method comprising:
receiving, at a data input of an analog computing system, an input data signal;
applying, by a computation unit having an input coupled with the data input, a weight to a signal received at its input to generate a weighted output signal; and
monitoring, by a controller, a parameter of the input signal and/or a parameter of the output signal and to controlling, by the controller, a level of accuracy of the weighted output signal based at least in part on the monitored parameter.