Systems and methods for sparsity exploiting

1. An integrated circuit, comprising:

a multi-layered neural network disposed in an analog multiplier array of a plurality of two-quadrant multipliers arranged in a memory sector of the integrated circuit, wherein at least one of the plurality of two-quadrant multipliers is a bias-free two-quadrant multiplier,

wherein each multiplier of the multipliers is wired to ground and draws a reduced amount of current when input signal values for input signals to transistors of the multiplier are near zero or zero, weight values of the transistors of the multiplier are near zero or zero, or a combination thereof, and

wherein each bias-free two-quadrant multiplier of the two-quadrant multipliers has a differential structure configured to allow programmatic compensation for overshoot if any one of two cells is set with a higher weight value than targeted.

2. The integrated circuit of claim 1 , wherein programmatic compensation for overshoot is performed by increasing a second weight value of a second cell of the two cells above an initially targeted weight value upon a determination that a first weight value of a first cell of the two cells is above a set value.

3. The integrated circuit of claim 1 , wherein sparsity in a neural network combined with a number of multipliers wired to ground reduces power consumption of the integrated circuit.

4. The integrated circuit of claim 1 , wherein each multiplier of the multipliers draws no current when the input signal values for the input signals to the transistors of the multiplier are zero, the weight values of the transistors of the multiplier are zero, or a combination thereof.

5. The integrated circuit of claim 1 , wherein the weight values correspond to synaptic weight values between neural nodes in the neural network disposed in the integrated circuit.

6. The integrated circuit of claim 5 , wherein input signal values multiplied by the weight values provide output signal values that are combined to arrive at a decision of the neural network.

7. The integrated circuit of claim 1 , wherein the transistor of the two-quadrant multipliers includes a metal-oxide-semiconductor field-effect transistor (“MOSFET”).

8. The integrated circuit of claim 1 , wherein the integrated circuit is configured for one or more application specific standard products (“ASSPs”) selected from keyword spotting, speaker identification, one or more audio filters, gesture recognition, image recognition, video object classification and segmentation, and autonomous vehicles including drones.

9. The integrated circuit of claim 1 , wherein the integrated circuit is configured to operate on battery power.

10. A method for compensation for overshoot in an integrated circuit, comprising:

training a multi-layered neural network disposed in an analog multiplier array of a plurality of two-quadrant multipliers in a memory sector of the integrated circuit, wherein at least one of the plurality of two-quadrant multipliers is a bias-free two-quadrant multiplier,

wherein each multiplier of the multipliers is wired to ground and draws a first amount of current when input signal values for input signals to transistors of the multiplier are near zero or zero, weight values of the transistors of the multiplier are near zero or zero, or a combination thereof; and

wherein each bias-free two-quadrant multiplier of the two-quadrant multipliers has a differential structure configured to allow programmatic compensation for overshoot if any one of two cells is set with a higher weight value than targeted.

11. The method of claim 10 , wherein the programmatic compensation for overshoot comprises:

determining whether a first weight value of a first cell of the two cells is above a set value; and

increasing a second weight value of a second cell of the two cells so that the second weight value is above an initially targeted weight value of the second cell.

12. The method of claim 10 , further comprising:

encouraging sparsity in the neural network by training with a training algorithm configured to reduce a plurality of the input signal values, the weight values, or the combination thereof, thereby enabling a reduced power consumption by the neuromorphic integrated circuit.

13. The method of claim 10 , wherein each multiplier of the multipliers draws no current when the input signal values for the input signals to the transistors of the multiplier are zero, the weight values of the transistors of the multiplier are zero, or a combination thereof.

14. The method of claim 10 , further comprising:

tracking rates of change for the weight values of each multiplier of the multipliers during the training;

determining if certain weight values are trending toward zero and how quickly those certain weight values are trending toward zero; and

reducing weight values during training to encourage sparsity in the neural network.

15. The method of claim 10 , wherein the training is an iterative training and the weight values are adjusted with each iteration of the training.

16. The method of claim 10 , wherein the weight values correspond to synaptic weight values between neural nodes in the neural network of the integrated circuit.

17. The method of claim 10 , further comprising:

minimizing a cost function that includes a quantity of non-zero weight values for the weight values with an optimization function including gradient descent, back-propagation, or both gradient descent and back-propagation, wherein an estimate of power consumption of the integrated circuit is used as a component of the cost function.

18. The method of claim 10 , further comprising:

setting a subset of the weight values to zero before training the neural network to encourage sparsity in the neural network.