Search

CN-121998010-A - Photon simulation iterative computing device

CN121998010ACN 121998010 ACN121998010 ACN 121998010ACN-121998010-A

Abstract

The photon simulation iterative computing device comprises a first optical signal, a second optical signal, a third optical signal, a fourth optical signal and a fifth optical signal, wherein the first optical signal is divided into the second optical signal, the third optical signal, the fourth optical signal and the fifth optical signal, the power of the second optical signal is modulated according to a first modulation voltage corresponding to an original value, a sixth optical signal is generated, the sixth optical signal is divided into a seventh optical signal and an eighth optical signal, the seventh optical signal is subjected to photon multiplication and addition calculation by using a second modulation voltage corresponding to a solving matrix, a ninth optical signal is generated, the eighth optical signal is used for generating a tenth optical signal corresponding to a proportional term, the fourth optical signal and the first modulation voltage are used for generating an eleventh optical signal corresponding to the original value, the fifth optical signal is used for generating a first output current by using a fifth optical signal and a first modulation voltage, the second output current corresponding to a new value is generated by using the first output current and the second output current.

Inventors

  • Hu Fangchen
  • ZOU PENG
  • ZHAO YIHENG
  • HE ZIQIANG
  • XU BO
  • CHU WEI
  • CAI HAIWEN

Assignees

  • 张江国家实验室

Dates

Publication Date
20260508
Application Date
20241106

Claims (14)

  1. 1. A photon simulation iterative computing device that performs iterative computation on a variable vector according to an iterative formula based on a solution matrix, comprising: the laser generation module outputs a first optical signal; A first power distribution module that equally divides the first optical signal into a second optical signal, a third optical signal, a fourth optical signal, and a fifth optical signal having the same power; the photoelectric modulation module modulates the power of the second optical signal according to a first modulation voltage corresponding to the original value of the variable vector to generate a sixth optical signal; a second power distribution module dividing the sixth optical signal equally into a seventh optical signal and an eighth optical signal having the same power; a photon multiplication and addition module for performing photon multiplication and addition calculation on the seventh optical signal by using a second modulation voltage corresponding to the solving matrix to generate a ninth optical signal, and An iteration module that generates a tenth optical signal corresponding to a proportional term in the iteration formula using the eighth optical signal, generates an eleventh optical signal corresponding to an original value term in the iteration formula using the fourth optical signal and the first modulation voltage, generates a first output current using the fifth optical signal by photoelectrically converting an optical domain accumulated value of the ninth optical signal, the tenth optical signal, and the eleventh optical signal, generates a second output current corresponding to a differential term in the iteration formula using the third optical signal and the first modulation voltage, and generates an output voltage corresponding to a new value of the variable vector using the first output current and the second output current.
  2. 2. The photon simulation iterative computing device of claim 1, wherein, The iterative module includes a differentiating unit that generates a twelfth optical signal using the first modulated voltage and the third optical signal, and generates the second output current using the twelfth optical signal, a regulating voltage, and a fourth modulated voltage corresponding to a momentum factor in the differential term.
  3. 3. The photon simulation iterative computing device of claim 2, The differentiating unit includes: a first phase modulator that modulates a phase of the third optical signal according to the first modulation voltage to generate the twelfth optical signal; a first power divider dividing the twelfth optical signal equally into thirteenth and fourteenth optical signals having the same power; a second power divider dividing the thirteenth optical signal equally into a fifteenth optical signal and a sixteenth optical signal having the same power; a second phase modulator for adjusting the delay and phase of the fifteenth optical signal according to the adjustment voltage to generate a seventeenth optical signal; An optical combiner that combines the sixteenth optical signal and the seventeenth optical signal into an eighteenth optical signal; A first photoelectric modulation unit for modulating the power of the eighteenth optical signal according to the fourth modulation voltage to generate a nineteenth optical signal, and And a first photoelectric conversion unit that couples the fourteenth optical signal and the nineteenth optical signal by multimode interference and performs photoelectric conversion, thereby generating the second output current.
  4. 4. A photon simulation iterative computing device according to any one of claim 1 to 3, The iteration module comprises: A multiplexing unit for performing optical domain summation of the ninth optical signal, the tenth optical signal and the eleventh optical signal to generate a twentieth optical signal, and And a second photoelectric conversion unit that couples the fifth optical signal and the twentieth optical signal by multimode interference and performs photoelectric conversion, thereby generating the first output current.
  5. 5. A photon simulation iterative computing device according to any one of claim 1 to 3, The iteration module comprises a scaling unit that modulates the power of the eighth optical signal according to a third modulation voltage corresponding to an annealing factor in the scaling term, generating the tenth optical signal.
  6. 6. A photon simulation iterative computing device according to any one of claim 1 to 3, The iteration module comprises a second photoelectric modulation unit which modulates the power of the fourth optical signal according to the first modulation voltage to generate the eleventh optical signal.
  7. 7. A photon simulation iterative computing device according to any one of claim 1 to 3, The iterative module comprises a transimpedance amplifier unit which converts the sum of the first output current and the second output current into a voltage and amplifies the voltage so as to obtain the output voltage.
  8. 8. The photon simulation iterative computing apparatus of claim 7, And adjusting the voltage amplification factor of the transimpedance amplifier unit to enable the new value of the variable vector to be in linear relation with the original value of the variable vector, so as to calibrate the iterative formula.
  9. 9. A photon simulation iterative computing device according to any one of claim 1 to 3, The iteration module further comprises an optical domain non-linear unit mapping the variable of the ninth optical signal to a binary or continuous time value using a non-linear function, thereby obtaining a twenty-first optical signal, The iteration module replaces the ninth optical signal with the twenty-first optical signal to generate the first output current.
  10. 10. A photon simulation iterative computing device according to any one of claim 1 to 3, The electric domain parts in the photon simulation iterative computation device are electrically connected through electric domain adjustable delay lines, and the electric domain adjustable delay lines enable the delay of the electric domain parts to be kept consistent.
  11. 11. A photon simulation iterative computing device according to any one of claim 1 to 3, Comprising N iteration modules, wherein N is a positive integer which is greater than 1 and corresponds to the dimension of the variable vector, The first power distribution module generates N second optical signals, N third optical signals, N fourth optical signals and N fifth optical signals, and provides the N third optical signals, the N fourth optical signals and the N fifth optical signals to the corresponding iteration modules respectively.
  12. 12. The photon simulation iterative computing apparatus of claim 11, The photoelectric modulation module comprises N third photoelectric modulation units, and each third photoelectric modulation unit modulates the power of a corresponding second optical signal in N second optical signals according to the first modulation voltage from the corresponding iteration module in the N iteration modules to generate N sixth optical signals.
  13. 13. The photon simulation iterative computing apparatus of claim 12, The second power distribution module comprises N third power distributors, each third power distributor uniformly divides each sixth optical signal into the seventh optical signal and the eighth optical signal, and each eighth optical signal is provided for the corresponding iteration module.
  14. 14. The photon simulation iterative computing device of claim 13, The solution matrix is an N x N matrix, The photon multiplication and addition module is composed of an array of N multiplied by N photoelectric conversion devices, the light transmittance of the photoelectric conversion devices corresponds to a scaling factor in the iteration formula, the photoelectric conversion devices modulate the power of the corresponding seventh optical signals according to the second modulation voltage respectively, and the modulated optical signals in the same column are subjected to optical domain accumulation, so that N ninth optical signals are generated and provided for the corresponding iteration modules respectively.

Description

Photon simulation iterative computing device Technical Field The invention relates to the field of photon simulation calculation, in particular to a photon simulation iterative calculation device for solving the optimization problem. Background The solution of the combinatorial optimization problem is widely applied to the scenes of logistics and supply chain management, production and manufacturing, financial investment combination optimization, network and communication, artificial intelligence learning and the like. Some of the more well-known problems include vehicle path planning problems, warehouse layout optimization problems, workshop scheduling problems, boxing problems, investment portfolio optimization problems and the like, and play a great role in social production and life. However, the combinatorial optimization problem is typically an NP problem, where the solution time and the problem size are exponentially related. For a problem of tens of thousands of scales, a traditional computer can even take a period of time in months to solve to obtain an optimal solution or a suboptimal solution. For larger scale problems, traditional computers have failed. For this reason, conventionally, a solution to the problem of combination optimization by optical computation has been proposed. Optical computing is an emerging analog computing approach. By taking light as a carrier wave and converting an analog electric signal into light through an electro-optical modulator to carry out numerical calculation, the optical calculation can realize the calculation requirements of low energy consumption, low time delay and high calculation power, and is particularly suitable for solving the combination optimization problem. Analog Iterative Machine (AIM) is a typical electro-optical hybrid computing architecture that combines analog electrical computing, and by effectively combining optical and analog electrical computing, AIM exhibits the advantages of low memory access and high bandwidth over conventional von neumann architectures. Fig. 5 shows a schematic diagram of a conventional AIM implementation, in which electrical signals are represented by dashed arrows and optical signals are represented by solid arrows. As shown in fig. 5, the overall system includes a Laser (LD), a plurality of optoelectronic modulators (EOMs), a photon matrix calculation unit (oMAC), a plurality of transimpedance amplifiers (TIAs), a plurality of adders (add), and a plurality of iteration modules. The photoelectric modulator modulates the electric signals x 1(t)~xN (t) corresponding to the calculation result of the iterative calculation from the iterative module into the laser signals emitted by the lasers and provides the laser signals to the photon matrix calculation unit. The photon matrix calculation unit combines a plurality of Photodiodes (PD) arranged at the rear end inside the photon matrix calculation unit, performs photon multiplication and addition calculation on the optical signals from the photoelectric modulator based on the solving matrix W, performs photoelectric conversion, and passes through a transimpedance amplifier and an adder at the rear end to calculateIs an electrical analog signal of (a)And respectively provided to each iteration module. The iteration module comprises a proportional unit consisting of an amplifier (amp.) and a proportional term (β (t)) and a differential unit consisting of a differentiator (device). The iteration module uses limiting amplifier (non linear) to simulate the electric signalNonlinear operation is performed, and the proportional term from the proportional unit, the differential term from the differential unit and the nonlinear linear matrix operation result from the limiting amplifier are summed by two summing circuits (add) to obtain the final calculation result Disclosure of Invention The invention aims to solve the technical problems Although the iterative delay of the prior AIM implementation is greatly reduced compared with the traditional computer, differential, proportion, nonlinear and other operations are still completed through electric calculation except matrix vector multiplication, so that the photoelectric conversion from optical calculation to analog electric calculation greatly reduces the energy consumption advantage of the AIM compared with the traditional computer. In addition, existing AIM implementations employ a large number of analog electrical devices that have bandwidth limited characteristics, which can further limit the latency advantages of analog computing systems. The invention aims to solve the technical problems, and aims to provide a photon simulation iterative computing device which can effectively improve the bandwidth of a system, thereby reducing the time delay of simulation iterative computing, improving the computing precision and remarkably reducing the power consumption of the system. Technical proposal for solving the technical problems Th