Neurosynaptic Core Utilization via Computational Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neurosynaptic networks require a large number of cores, which increases costs and energy consumption, as each core contributes significantly to the overall system costs and power usage, despite the number of neurons and axons being less costly.
Innovation Solution
The method involves dividing large computational blocks into smaller subunits and reallocating them across a reduced number of cores, increasing the number of neurons while maintaining or slightly modifying spike activity, thereby reducing the total number of cores needed without affecting output functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If a large number of cores are used in neurosynaptic networks, then computational capacity and processing power are improved, but system cost and energy consumption increase significantly
Solution Approach 1:
The patent divides computational blocks into smaller subunits that can be distributed across fewer cores. By segmenting the computational workload and reorganizing how neurons and axons are allocated across cores, the system achieves the same computational capacity with reduced core count, thereby lowering energy consumption while maintaining processing power
Solution Approach 2:
The patent merges computational functionality by increasing the density of neurons and axons within each core. By combining more neural elements into fewer cores through optimized allocation, the system maintains overall computational capacity while reducing the total number of cores required, thus decreasing energy consumption
2Power
If a large number of cores are used in neurosynaptic networks, then computational capacity is improved, but system cost increases
Solution Approach 1:
By segmenting computational blocks and redistributing neural elements across fewer cores, the patent reduces the total core count required. This segmentation strategy maintains computational capacity while simplifying the overall system architecture, leading to reduced manufacturing costs and lower system complexity
Solution Approach 2:
The patent changes key parameters by increasing neuron and axon density within each core while reducing the total number of cores. This parameter transformation allows the system to achieve the same computational output with fewer physical components, thereby reducing device complexity and associated costs
3Productivity
If functional units are divided into subunits and reallocated, then core utilization is optimized and number of cores is reduced, but network configuration complexity increases
Solution Approach 1:
The patent applies segmentation by dividing functional units into subunits and redistributing them across cores. While this improves core utilization, the patent manages the accompanying configuration complexity through systematic reorganization methods that maintain network functionality while optimizing resource allocation
Data Source
AI summary
Core utilization optimization by dividing computational blocks across neurosynaptic cores is provided. In some embodiments, a neural network description describing a neural network is read. The neural network comprises a plurality of functional units on a plurality of cores. A functional unit is selected from the plurality of functional units. The functional unit is divided into a plurality of subunits. The plurality of subunits are connected to the neural network in place of the functional unit. The plurality of functional units and the plurality of subunits are reallocated between the plurality of cores. One or more unused cores are removed from the plurality of cores. An optimized neural network description is written based on the reallocation.