METHOD AND APPARATUS FOR UNSUPERVISED TRAINING OF INPUT SYNAPSES OF PRIMARY VISUAL CORTEX SIMPLE CELLS AND OTHER NEURAL CIRCUITS

• US 20120303566A1
• Filed: 05/25/2011
• Published: 11/29/2012
• Est. Priority Date: 05/25/2011
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Aparin, Vladimir
• Granted Patent: US 8,583,577 B2
• Time in Patent Office: Days
• US Class Current: 706/25
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

Certain aspects of the present disclosure present a technique for unsupervised training of input synapses of primary visual cortex (V1) simple cells and other neural circuits. The proposed unsupervised training method utilizes simple neuron models for both Retinal Ganglion Cell (RGC) and V1 layers. The model simply adds the weighted inputs of each cell, wherein the inputs can have positive or negative values. The resulting weighted sums of inputs represent activations that can also be positive or negative. In an aspect of the present disclosure, the weights of each V1 cell can be adjusted depending on a sign of corresponding RGC output and a sign of activation of that V1 cell in the direction of increasing the absolute value of the activation. The RGC-to-V1 weights can be positive and negative for modeling ON and OFF RGCs, respectively.

Claims

1. An electrical circuit, comprising:

a plurality of Retinal Ganglion Cell (RGC) circuits, wherein each of the RGC circuits generates, at an output, a sum of weighted inputs from receptor circuits associated with that RGC circuit;

a plurality of primary visual cortex cell (V1) circuits, wherein each of the V1 circuits generates another sum of weighted outputs of a subset of the RGC circuits; and

a circuit configured to adjust weights applied on the outputs for generating the other sum, wherein

the adjustment of one of the weights is based on at least one of one of the outputs on which that weight is applied or the other sum.

2. The electrical circuit of claim 1, wherein the adjustment of that weight is based on a sign of that output and a sign of the other sum.

3. The electrical circuit of claim 1, wherein the adjustment of that weight is based on a sign of that output and a value of the other sum.

4. The electrical circuit of claim 1, wherein:

that weight is increased, if that output and the other sum are both positive or both negative, and

that weight is decreased, if a sign of that output and a sign of the other sum are not same.

5. The electrical circuit of claim 1, wherein that V1 circuit outputs a signal, if the other sum generated by that V1 circuit exceeds a threshold.

6. The electrical circuit of claim 1, wherein the circuit is also configured to adjust each of the weights by performing a binary shift operation.

7. The electrical circuit of claim 1, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.

8. A method for implementing a neural system, comprising:

generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit;

generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; and

adjusting weights applied on the outputs for generating the other sum,

wherein the adjustment of one of the weights is based on at least one of one of the outputs on which that weight is applied or the other sum.

9. The method of claim 8, wherein the adjustment of that weight is based on a sign of that output and a sign of the other sum.

10. The method of claim 8, wherein the adjustment of that weight is based on a sign of that output and a value of the other sum.

11. The method of claim 8, wherein:

that weight is increased, if that output and the other sum are both positive or both negative, and

that weight is decreased, if a sign of that output and a sign of the other sum are not same.

12. The method of claim 8, further comprising:

generating a signal at an output of that V1 circuit, if the other sum generated by that V1 circuit exceeds a threshold.

13. The method of claim 8, wherein adjusting each of the weights comprises:

performing a binary shift operation.

14. The method of claim 8, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.

15. An apparatus, comprising:

means for generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the apparatus, a sum of weighted inputs from receptor circuits associated with that RGC circuit;

means for generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the apparatus, another sum of weighted outputs of a subset of the RGC circuits; and

means for adjusting weights applied on the outputs for generating the other sum,

wherein the adjustment of one of the weights is based on at least one of one of the outputs on which that weight is applied or the other sum.

16. The apparatus of claim 15, wherein the adjustment of that weight is based on a sign of that output and a sign of the other sum.

17. The apparatus of claim 15, wherein the adjustment of that weight is based on a sign of that output and a value of the other sum.

18. The apparatus of claim 15, wherein:

that weight is increased, if that output and the other sum are both positive or both negative, and

that weight is decreased, if a sign of that output and a sign of the other sum are not same.

19. The apparatus of claim 15, further comprising:

means for generating a signal at an output of that V1 circuit, if the other sum generated by that V1 circuit exceeds a threshold.

20. The apparatus of claim 15, wherein the means for adjusting each of the weights comprises:

means for performing a binary shift operation.

21. The apparatus of claim 15, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.