Method to perform convolutions between arbitrary vectors using clusters of weakly coupled oscillators

Patent metadata

Field	Value
Publication number	US-9998130-B2
Application number	US-201615203596-A
Country	US
Kind code	B2
Filing date	Jul 6, 2016
Priority date	Jul 6, 2016
Publication date	Jun 12, 2018
Grant date	Jun 12, 2018

Abstract

A method to perform convolutions between arbitrary vectors includes estimating a first degree of match for a difference between a first vector having a plurality of first elements and a second vector having a plurality of second elements using a first cluster of coupled oscillators, estimating a second degree of match for the first vector using a second cluster of coupled oscillators, estimating a third degree of match for the second vector using a third cluster of coupled oscillators, deriving a first squared L 2 norm from the first degree of match, deriving a second squared L 2 norm from the second degree of match, deriving a third squared L 2 norm from the third degree of match, adding the second squared L 2 norm and the third squared L 2 norm, and subtracting the first squared L 2 norm to form a sum, and dividing the sum by two.

First claim

What is claimed is:

1. A method to perform convolutions between arbitrary vectors comprising:
computing a first degree of match for a difference between a first vector having a plurality of first elements and a second vector having a plurality of second elements using a first cluster of weakly coupled oscillators;
computing a second degree of match for the first vector using a second cluster of weakly coupled oscillators;
computing a third degree of match for the second vector using a third cluster of weakly coupled oscillators;
estimating a first squared L 2 norm from the first degree of match;
estimating a second squared L 2 norm from the second degree of match;
estimating a third squared L 2 norm from the third degree of match; and
adding the second squared L 2 norm and the third squared L 2 norm, and subtracting the first squared L 2 norm to form a sum.

2. The method of claim 1 wherein:
estimating the first squared L 2 norm from the first degree of match comprises using a first precomputed lookup table or a first precomputed piecewise linear function for deriving the first squared L 2 norm from the first degree of match;
estimating the second squared L 2 norm from the second degree of match comprises using a second precomputed lookup table or a second precomputed piecewise linear function for deriving the second squared L 2 norm from the second degree of match; and
estimating the third squared L 2 norm from the third degree of match comprises using a third precomputed lookup table or a third precomputed piecewise linear function for deriving the third squared L 2 norm from the third degree of match.

3. The method of claim 1 wherein:
estimating the first squared L 2 norm from the first degree of match comprises:
characterizing the first degree of match for different first squared L 2 norms using the first cluster of weakly coupled oscillators to estimate a first degree of match for a plurality of samples of elements of the first vector and for a plurality of samples of elements of the second vector to form a correspondence between a respective first degree of match and a respective first squared L 2 norm, wherein each element of the first vector and each element of the second vector ranges between −1 and +1;

estimating the second squared L 2 norm from the second degree of match comprises:
characterizing the second degree of match for different second squared L 2 norms using the second cluster of weakly coupled oscillators to estimate a second degree of match for a plurality of samples of elements of the first vector to form a correspondence between a respective second degree of match and a respective second squared L 2 norm, wherein each element of the first vector ranges between −1 and +1; and

deriving the third squared L 2 norm from the third degree of match comprises:
characterizing the third degree of match for different third squared L 2 norms using the third cluster of weakly coupled oscillators to estimate a third degree of match for a plurality of samples of elements of the second vector to form a correspondence between a respective third degree of match and a respective third squared L 2 norm, wherein each element of the second vector ranges between −1 and +1.

4. The method of claim 3 :
wherein estimating the first degree of match for the difference between the first vector having a plurality of first elements and the second vector having a plurality of second elements comprises estimating the first degree of match for a difference between the first vector and the second vector, wherein each element of the first vector and each element of the second vector ranges between −1 and +1;
wherein estimating the second degree of match for the first vector comprises estimating the second degree of match wherein each element of the first vector ranges between −1 and +1; and
wherein estimating the third degree of match for the second vector comprises estimating the third degree of match wherein each element of the second vector ranges between −1 and +1; and
further comprising
calculating a first L 1 norm of the first vector;
calculating a second L 1 norm of the second vector;
adding the second squared L 2 norm and the third squared L 2 norm, and subtracting the first squared L 2 norm, a first scaling factor times the first L 1 norm, a second scaling factor times the second L 2 norm, and a third factor to form a sum; and
dividing the sum by two.

5. The method of claim 1 :
wherein the second cluster of weakly coupled oscillators is the first cluster of weakly coupled oscillators; and
wherein the third cluster of weakly coupled oscillators is the first cluster of weakly coupled oscillators.

6. The method of claim 1 further comprising:
estimating a plurality of first degree of matches for a plurality of first vectors and a plurality of second vectors over a range of values of the first elements of the first vectors and over a range of values of the second elements of the second vectors to characterize a relationship between a respective first degree of match and a respective first squared L 2 norm;
estimating a plurality of second degree of matches for a plurality of first vectors over the range of values of the first elements of the first vectors to characterize a relationship between a respective second degree of match and a respective second squared L 2 norm; and
estimating a plurality of third degree of matches for a plurality of second vectors over the range of values of the second elements of the second vectors to characterize a relationship between a respective third degree of match and a respective third squared L 2 norm.

7. The method of claim 1 wherein:
estimating the second degree of match for the first vector using the second cluster of weakly coupled oscillators comprises estimating a second degree of match for a difference between the first vector and a third vector having a plurality of zero elements; and
estimating the third degree of match for the second vector using the third weakly cluster of coupled oscillators comprises estimating the third degree of match for a difference between the second vector and the third vector having the plurality of zero elements.

8. The method of claim 1 wherein:
the first cluster of weakly coupled oscillators comprises a resonant body oscillator or a spin torque oscillator;
the second cluster of weakly coupled oscillators comprises a resonant body oscillator or a spin torque oscillator; and
the third cluster of weakly coupled oscillators comprises a resonant body oscillator or a spin torque oscillator.

9. The method of claim 1 wherein:
the first cluster of weakly coupled oscillators comprises:
a plurality of difference circuits, each respective difference circuit coupled to a respective element of the first vector and to a respective element of the second vector and having a respective difference output;
a plurality of oscillators, each respective oscillator having a respective input coupled to a respective difference output, and having a respective oscillator output;
a summer for adding the respective oscillator outputs; and
an integrator for integrating over a plurality of cycles of the oscillators and having an integrator output for providing an estimate of the first degree of match.

10. The method of claim 9 wherein:
each difference circuit comprises a transconductance amplifier; and
each oscillator comprises a time encoded oscillator comprising:
a summer coupled to the difference circuit;
an integrator coupled to the summer;
a hysteresis quantizer coupled to the integrator; and
a 1-bit digital to analog converter coupled to the hysteresis quantizer having an out

Assignees

• Hrl Lab Llc
• Univ Of Pittsburgh—Of The Commonwealth System Of Higher Education
• Balazs Anna
• Univ Of Pittsburg—Of The Commonwealth System Of Higher Education

Inventors

• Pilly Praveen K
• Cruz-Albrecht Jose
• Srinivasa Narayan
• Levitan Steven P
• Chiarulli Donald M

Classifications

• G06N3/065

  Analogue means · CPC title

• G06T5/20

  using local operators · CPC title

• H04L9/0838

  Key agreement, i.e. key establishment technique in which a shared key is derived by parties as a function of information contributed by, or associated with, each of these (network architectures or network communication protocols for key exchange in a packet data network H04L63/061) · CPC title

• H04L9/12

  Transmitting and receiving encryption devices synchronised or initially set up in a particular manner · CPC title

• H03K3/354

  Astable circuits · CPC title