Neural Network Resource Estimation via Structural Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating the resource amount of dedicated processing resources required for emerging technologies like deep learning, high-performance computing, and artificial intelligence are inefficient, often relying on prior knowledge that users lack, leading to over-estimation and poor resource utilization.
Innovation Solution
A method and device that obtain a structural representation of a neural network to determine the required resource amount for deep learning processing, allowing for accurate prediction and scheduling of dedicated processing resources without requiring extensive user knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods relying on prior knowledge are used to estimate resource amounts, then users can request dedicated processing resources, but the estimation is inefficient and leads to over-estimation causing poor resource utilization
Solution Approach 1:
The system automatically extracts structural representation from the neural network model and computes resource requirements without requiring users to manually provide prior knowledge or configuration information. The deep learning model autonomously analyzes the network structure, layer attributes, and computational patterns to self-determine accurate resource amounts, eliminating the need for user expertise while improving estimation accuracy and reducing resource waste
Solution Approach 2:
The patent transforms the resource estimation problem by changing from manual parameter input to automated structural analysis. Instead of relying on users to provide prior knowledge about resource requirements, the system extracts structural representation parameters from the neural network model itself (such as layer types, neuron counts, activation functions) and uses these as input features to predict resource needs, fundamentally changing how resource amounts are determined
2Measurement precision
If users provide manual configuration information for resource estimation, then estimation can be performed, but it requires extensive user knowledge and increases operational complexity
Solution Approach 1:
The system automatically extracts structural representation from the neural network model and computes resource requirements without requiring users to manually provide prior knowledge or configuration information. The deep learning model autonomously analyzes the network structure, layer attributes, and computational patterns to self-determine accurate resource amounts, eliminating the need for user expertise while improving estimation accuracy and reducing resource waste
Solution Approach 2:
The patent introduces an intermediary deep learning model that acts as a mediator between the neural network structure and resource estimation. This intermediary automatically processes the structural representation, extracts relevant features, and computes resource requirements, shielding users from the complexity of resource calculation while maintaining high estimation accuracy without requiring extensive user knowledge
Data Source
AI summary
Embodiments of the present disclosure relate to a method, device and computer program product for determining a resource amount of dedicated processing resources. The method comprises obtaining a structural representation of a neural network for deep learning processing, the structural representation indicating a layer attribute of the neural network that is associated with the dedicated processing resources; and determining the resource amount of the dedicated processing resources required for the deep learning processing based on the structural representation. In this manner, the resource amount of the dedicated processing resources required by the deep learning processing may be better estimated to improve the performance and resource utilization rate of the dedicated processing resource scheduling.