STORAGE DEVICE WITH CONFIGURABLE NEURAL NETWORKS

• US 20160247080A1
• Filed: 02/19/2015
• Published: 08/25/2016
• Est. Priority Date: 02/19/2015
• Status: Active Grant

• Current Assignee: Seagate Technology LLC (Seagate Technology Holdings Plc)
• Original Assignee: Seagate Technology LLC (Seagate Technology Holdings Plc)
• Inventors: Younger, Scott, Trantham, Jon, Gomez, Kevin Arthur, Dropps, Frank, Khoueir, Antoine
• Granted Patent: US 11,295,202 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

An apparatus comprises a mass storage unit and a plurality of circuit modules including a machine learning module, a programmable state machine module, and input/output interfaces. Switching circuitry is configured to selectively couple the circuit modules. Configuration circuitry is configured to access configuration data from the mass storage unit and to operate the switching circuitry to connect the circuit modules according to the configuration data.

Claims

1. An apparatus comprising:

a mass storage unit;

a plurality of circuit modules including a machine learning module, a programmable state machine module, and input/output interfaces;

switching circuitry configured to selectively couple the circuit modules; and

configuration circuitry configured to access configuration data from the mass storage unit and to operate the switching circuitry to connect the circuit modules according to the configuration data.

2. The apparatus of claim 1, wherein the input/output interfaces comprise analog interfaces.

3. The apparatus of claim 1, wherein circuit modules further include a vector matrix multiplier module.

4. The apparatus of claim 1, wherein the machine learning module comprises a neural network module.

5. The apparatus of claim 1, wherein the machine learning module comprises a hidden Markov Model module.

6. The apparatus of claim 1, wherein the machine learning module comprises a scale-invariant feature transform module.

7. The apparatus of claim 1, wherein circuit modules further include a Viterbi module.

8. The apparatus of claim 1, wherein the switching circuitry is further configured to selectively couple the circuit modules to a portion of the mass storage unit.

9. A system comprising:

a storage unit;

a plurality of circuit modules including a machine learning module, a programmable state machine module, and input/output interfaces, the circuit modules configured to process a data stream and write output of such to the storage unit;

switching circuitry configured to selectively couple the circuit modules; and

configuration circuitry configured to access configuration data from the storage unit and to operate the switching circuitry to connect the circuit modules according to the configuration data.

10. The system of claim 9, wherein the data stream is a live stream of data received at a data input of the apparatus.

11. The system of claim 9, wherein the data stream is formed from content previously stored on the storage unit.

12. The system of claim 9, wherein the configuration circuitry is configured to store current configuration data using scan chains.

13. The system of claim 12, wherein the configuration circuitry is configured to store the current configuration data to the storage unit by capturing the current configuration data from the scan chains.

14. The system of claim 9, wherein the configuration circuitry is configured to:

receive a shut-down command; and

apply current configuration data to scan chains upon receipt of the shut-down command.

15. A method comprising:

receiving, from a host, a configuration defining connections between circuit modules of a storage compute device, the circuit modules including a machine learning module, a programmable state machine module, and input/output interfaces;

applying the configuration to switching circuitry to selectively couple the circuit modules; and

processing a data stream via the coupled circuit modules.

16. The method of claim 15, wherein the machine learning module comprises at least one of a neural network module, a hidden Markov Model module, and a scale-invariant feature transform module.

17. The method of claim 15, wherein the switching circuitry is further configured to selectively couple the circuit modules to a portion of a mass storage unit of the storage compute device.

18. The method of claim 15, further comprising writing an output of the processed data stream to a mass storage unit of the storage compute device.

19. The method of claim 18, wherein the data stream is a live stream of data received at a data input of the storage compute device.

20. The method of claim 18, wherein the data stream is formed from content previously stored on the mass storage unit.