Neural Network Initialization via Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining the structure and initial values of parameters in artificial neural networks is challenging, leading to time-consuming training and inaccurate outputs due to random initialization methods.
Innovation Solution
Efficient clustering of large-scale datasets into multiple clusters to determine the neural network structure and training samples, allowing for representative data-driven initialization of model parameters, reducing computational complexity and improving prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional random initialization methods are used for neural network parameters, then the neural network can be constructed, but the training process becomes time-consuming and the output accuracy is poor
Solution Approach 1:
The patent applies preliminary action by performing clustering analysis on the training data before neural network training to determine the optimal number of hidden layers and initialize parameters. This preliminary structuring of the network based on data characteristics eliminates the need for extensive trial-and-error training, directly reducing training time while improving output accuracy through data-driven parameter initialization.
2Measurement precision
If the neural network structure is determined without data analysis, then the network can be constructed quickly, but the structure does not match the training data characteristics leading to poor performance
Solution Approach 1:
The patent applies parameter changes by using clustering analysis to dynamically determine neural network structural parameters (number of hidden layers, nodes per layer) based on the actual characteristics of the training data. Instead of using fixed or randomly determined parameters, the system adjusts these structural parameters according to data distribution, improving prediction accuracy while the automated clustering process keeps the determination complexity manageable.
3Reliability
If clustering is performed on large-scale datasets using traditional methods, then complete clustering can be achieved, but the computational complexity is high and processing time is excessive
Solution Approach 1:
The patent applies segmentation by dividing the large-scale dataset clustering task into multiple iterative clustering operations. Instead of performing one comprehensive clustering on the entire dataset, the system performs sequential clustering iterations that progressively refine cluster assignments, maintaining clustering completeness while reducing the computational burden of each individual clustering operation, thereby improving overall processing speed.
Data Source
AI summary
In some aspects, systems and methods for efficiently clustering a large-scale dataset for improving the construction and training of machine-learning models, such as neural network models, are provided. A dataset used for training a neural network model configured can be clustered into a first set of clusters and a second set of clusters. The neural network model can be constructed with a number of nodes in a hidden layer that is based on the number of clusters in the first set of clusters. The neural network can be trained based on training samples selected from the second set of clusters. In some aspects, the trained neural network model can be utilized to satisfy risk assessment queries to compute output risk indicators for target entities. The output risk indicator can be used to control access to one or more interactive computing environments by the target entities.