Boltzmann Machine Circuit Using Thermal Noise for Local Minimum Escape
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Boltzmann machines face challenges in efficiently changing the amplitude of random noise due to large circuit scales and random noise generation, making it difficult to prevent solutions from falling into local minimum values.
Innovation Solution
The implementation of a Boltzmann machine with circuit units that include adders, comparison units, and digital arithmetic units capable of generating weighted input signals, using thermal noise and DA converters to vary the amplitude of random noise, allowing for a small-scale circuit to control the probability of output changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a random noise generator is used to change the probability of lattice output, then the temperature parameter can be adjusted, but the circuit scale becomes large and the noise amplitude control becomes difficult
Solution Approach 1:
The patent changes the parameter representation from direct analog noise amplitude control to digital domain control. By converting the temperature parameter to digital form and using digital-to-analog conversion only where necessary, the system achieves parameter adjustability while reducing circuit complexity. The digital random number generator and digital arithmetic operations replace the need for large-scale analog noise generation circuits.
2Temperature
If conventional analog noise generation is used, then temperature control is possible, but the noise amplitude cannot be appropriately changed due to random magnitude
Solution Approach 1:
The patent replaces the analog mechanical noise generation system with a digital system. Instead of using analog circuits to generate and control noise amplitude, the invention uses digital random number generation combined with digital arithmetic operations. The digital-to-analog converter is used only at the final stage where precise amplitude control is needed, allowing easy adjustment of noise amplitude through digital parameter setting rather than complex analog circuit tuning.
3Reliability
If simulated annealing is implemented with conventional circuits, then local minimum prevention is achieved, but the circuit scale increases
Solution Approach 1:
The patent segments the simulated annealing function into modular digital components: a digital random number generator, digital arithmetic units for weighted sum calculation, and digital comparison units. Each circuit unit processes one lattice point independently, allowing the system to maintain reliability for local minimum prevention while keeping individual circuit blocks small and manageable. The modular structure enables scalable implementation without proportionally increasing overall circuit complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables the effective variation of random noise amplitude with a small-scale circuit, enhancing the ability to escape local minima and achieve optimal solutions by stochastically changing the energy state.
Implementation Method 1
a first comparator that compares a thermal noise with a reference voltage to output a binary digital random signal
Implementation Method 2
a first DA converter that converts the digital random signal to an analog random signal, and varies a magnitude of the analog random signal
Implementation Method 3
a second comparator that input the analog random signal as the threshold signal, and compares the output signal of the adder with the threshold signal to generate the binary output signal with a predetermined probability function
Data Source
AI summary
Boltzmann machine includes a plurality of circuit units each having an adder that adds weighted input signals and a comparison unit that compares an output signal of the adder with a threshold signal to output a binary output signal; and digital arithmetic units each generating the weighted input signals by weighting the binary output signal of the circuit units with a weight. The comparison unit has a first comparator that compares a thermal noise with a reference voltage to output a binary digital random signal, a DA converter that converts the digital random signal to an analog random signal and varies a magnitude of the analog random signal, and a second comparator that compares the output signal of the adder with the analog random signal to generate the binary output signal with a predetermined probability function.