Neurosynaptic Network Optimization via Neuron Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neurosynaptic networks require a large number of neurons and axons, leading to increased size, cost, and energy consumption, which is a significant challenge in implementing efficient and scalable systems, especially in energy-sensitive applications.
Innovation Solution
The method involves optimizing neurosynaptic networks by reducing the number of neurons and axons through modifications in network topology, such as removing active neurons and axons, rearranging splitter neurons, and collapsing synapses, while ensuring that the network functionality remains unchanged.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of neurons and axons is increased to maintain network functionality, then the network's computational capability is improved, but the system size, cost, and energy consumption increase
Solution Approach 1:
The patent extracts and removes redundant neurons and axons from the neurosynaptic network through systematic analysis of network functionality. By identifying and eliminating unnecessary computational elements while preserving essential functions, the network achieves reduced size and lower resource requirements without sacrificing computational capability.
Solution Approach 2:
The patent merges multiple neurons into functional units that can perform equivalent computational tasks with fewer elements. By combining redundant computational functions into unified neural structures, the network maintains its processing power while reducing the total number of neurons and axons required.
2Productivity
If more neurons and axons are used to ensure network functionality, then the computational power is maintained, but the energy consumption increases
Solution Approach 1:
The patent identifies and removes energetically expensive but functionally redundant neurons and axons from the network. By extracting unnecessary computational elements, the system reduces its energy consumption profile while maintaining the computational power required for its intended functions.
Solution Approach 2:
The patent changes the structural parameters of the network by optimizing the configuration and connectivity of remaining neurons after reduction. By adjusting network topology and synaptic connections, the system achieves efficient energy utilization while preserving computational capabilities.
3Quantity of substance
If the number of cores is reduced to lower cost and energy consumption, then the system becomes more compact and efficient, but the network functionality may be compromised
Solution Approach 1:
The patent merges multiple core functions into fewer, more efficient cores by consolidating neural processing tasks. Through functional integration and shared resource utilization, the reduced number of cores maintains network functionality while achieving cost and energy efficiencies.
Solution Approach 2:
The patent designs reduced cores with multi-functional capabilities that can perform multiple computational tasks. By creating universal processing units that handle diverse neural operations, the system maintains full network functionality with fewer specialized cores.
Data Source
AI summary
Reduction in the number of neurons and axons in a neurosynaptic network while maintaining its functionality is provided. A neural network description describing a neural network is read. One or more functional unit of the neural network is identified. The one or more functional unit of the neural network is optimized. An optimized neural network description is written based on the optimized functional unit.