Ising Unit Neuron Selection Circuit for Optimization Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Ising unit implementations for solving multivariable optimization problems face inefficiencies in calculation speed due to limited state transitions, even when performed in parallel, as many neurons do not change states, leading to suboptimal processing speed.
Innovation Solution
An information processing apparatus with a plurality of Ising units, each including a calculation circuit for updating energy values and a state transition determination circuit to assess possible state changes, along with an updated neuron selection circuit that identifies and selects neurons with possible state transitions, enhancing the probability of state changes and improving calculation speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallel processing is performed with multiple Ising units, then processing capacity is improved, but calculation speed is not appreciably improved because many selected neurons do not make state transitions
Solution Approach 1:
The patent applies preliminary action by pre-calculating the energy change amount for each neuron before selecting which neuron to update. The selection circuit chooses a neuron based on having a non-zero energy change amount, ensuring that the selected neuron will actually transition state. This prevents the waste of parallel processing cycles on neurons that would not change state, thereby improving calculation speed while maintaining processing capacity.
2Ease of operation
If random neuron selection is used in parallel processing, then implementation simplicity is maintained, but efficiency decreases due to selection of neurons that do not transition
Solution Approach 1:
The patent implements feedback by using the calculated energy change amount as a selection criterion. The selection circuit receives information about which neurons have non-zero energy change amounts and uses this feedback to guide the selection process. This ensures that only neurons capable of state transition are selected, improving processing efficiency without significantly complicating the implementation.
3Reliability
If all neurons are updated in parallel, then convergence is maintained, but calculation speed deteriorates due to unnecessary updates
Solution Approach 1:
The patent applies the taking out principle by extracting only the neurons that have non-zero energy change amounts from the set of all neurons. Instead of updating all neurons in parallel, the system identifies and updates only those neurons that will actually contribute to state transition. This reduces unnecessary updates and improves calculation speed while maintaining convergence through the selective update of relevant neurons.
Data Source
AI summary
Each energy value calculation circuit calculates an energy value, based on a sum total of values obtained by multiplying state values of a plurality of second neurons coupled with a first neuron by corresponding weighting values indicating coupling strengths, and updates the energy value, based on identification information of an updated neuron whose state is updated among the plurality of second neurons and a state value of the updated neuron. Each state transition determination circuit outputs, based on a second energy value and a noise value, a determination signal indicating a determination result of whether a change in a state value of the first neuron is possible. An updated neuron selection circuit selects, based on received determination signals, a first neuron a change in whose state value is possible and outputs identification information of the selected first neuron as identification information of the updated neuron.