Systems and methods for modeling binary synapses
View Patent ↗Methods and system for modeling the behavior of binary synapses are provided. In one aspect, a method of modeling synaptic behavior includes receiving an analog input signal and transforming the analog input signal into an N-bit codeword, wherein each bit of the N-bit codeword is represented by an electronic pulse. The method includes loading the N-bit codeword into a circular shift register and sending each bit of the N-bit codeword through one of N switches. Each switch applies a corresponding weight to the bit to produce a weighted bit. A signal corresponding to a summation of the weighted bits is output and represents a synaptic transfer function characterization of a binary synapse.
1. A method comprising:
receiving an analog input signal;
transforming the analog input signal into an N-bit codeword, wherein each bit of the N-bit codeword is represented by an electronic pulse, wherein N is greater than one;
loading the N-bit codeword into an N-bit circular shift register, wherein each bit of the N-bit circular shift register is coupled to a corresponding switch of N switches;
sending each bit of the N-bit codeword from the N-bit circular shift register through the corresponding switch of the N switches, wherein each switch of the N switches applies a corresponding weight to the bit to produce a weighted bit;
summing the N weighted bits using an integrator to obtain a summation of the N weighted bits; and
outputting a signal corresponding to the summation of the N weighted bits produced by the N switches, the signal representing a synaptic transfer function.
2. The method of claim 1 , wherein transforming the input signal into an N-bit codeword further comprises each pulse representing a bit as a high voltage or a low voltage of an electronic signal.
3. The method of claim 1 , wherein loading the N-bit codeword into the N-bit circular shift register further comprises loading the N bits of the N-bit codeword in parallel into the N-bit circular shift register.
4. The method of claim 1 , further comprising:
circularly shifting each bit value of the N-bit codeword through each bit of the N-bit circular shift register.
5. The method of claim 4 further comprising:
after each circular shift of the N-bit circular shift register, sending each shifted bit of the N-bit codeword from the N-bit circular shift register through one of the N switches.
6. The method of claim 5 further comprising:
accumulating weighted bits after each circular shift of the N-bit circular shift register.
7. The method of claim 1 further comprising updating the weights of each switch prior to sending the N bits of the N-bit codeword.
8. The method of claim 1 , wherein sending each bit of the N-bit codeword through one of N switches further comprises sending the N-bits through the N switches in parallel.
9. The method of claim 1 , wherein each switch further comprises a memristor crossbar junction of a crossbar array.
10. The method of claim 9 , wherein the crossbar array comprises:
a first set of approximately parallel nanowires, wherein each nanowire in the first set is coupled to one bit of the N-bit circular shift register;
a second set of approximately parallel nanowires, wherein each nanowire in the second set overlaps the nanowires in the first set.
11. The method of claim 1 , wherein the integrator comprises a capacitor to store the sums of the weighted bits as electric charge.
12. The method of claim 1 , wherein the integrator comprises a differential amplifier that receives each bit at an input.
13. The method of claim 1 , wherein transforming the analog input signal into the N-bit codeword comprises using a thermometer encoding.
14. A system for modeling binary synaptic behavior comprising:
a first set of approximately parallel nanowires, wherein each nanowire receives a pulse corresponding to a bit of an N-bit codeword;
a second set of approximately parallel nanowires, wherein each nanowire in the second set overlaps the nanowires in the first set;
a memristor located at each nanowire intersection, each memristor connects a nanowire in the first set to a nanowire in the second set and applies a weight to a bit of the N-bit codeword to produce a weighted bit; and
at least one integrator, each integrator electronically connected to a nanowire in the second set and sums the weighted bits to produce a output representing a synaptic transfer function.
15. The system of claim 14 further comprising a circular shift register electronically connected to the first set of nanowires, the circular shift register circularly shifts the bit values of the N-bit codeword and sends each shifted bit through one of the N switches.
16. The system of claim 14 , wherein each pulse further comprises a high voltage or a low voltage of an electronic signal.
17. The system of claim 14 , wherein the integrator further comprises a capacitor that stores the sums of the weighted bits as electric charge.
18. The system of claim 14 , wherein the integrator further comprises a differential amplifier that receives each bit at an input.
19. The system of claim 14 further comprising an analog-to-digital converter that converts an analog signal into the N-bit codeword.