Neural Network Representation Learning for Complex Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional clustering methods, such as those using infinite Gaussian mixture models, deteriorate in performance for complex data that cannot be represented by a Gaussian distribution.
Innovation Solution
A learning method involving a neural network-based approach that inputs data, generates representation data, clusters using an infinite mixture Gaussian distribution, calculates an evaluation scale for clustering performance, and learns neural network parameters to improve clustering accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional clustering methods using infinite Gaussian mixture models are used, then the clustering process is simple and automatic, but the clustering performance deteriorates for complex data that cannot be represented by a Gaussian distribution
Solution Approach 1:
The patent introduces representation data as an intermediary between the original complex data and the clustering process. The neural network converts original data into representation data that captures essential features while being suitable for Gaussian mixture model clustering, thus mediating between complex data structures and simple clustering algorithms
Solution Approach 2:
The patent replaces the direct application of Gaussian mixture models on complex data with a neural network-based representation learning system. The neural network acts as a feature extractor that transforms complex data into a form suitable for traditional clustering, substituting mechanical clustering directly on raw data with a two-stage process involving representation learning
2Measurement precision
If a neural network-based representation learning approach is introduced, then clustering performance on complex data improves, but the system complexity and computational requirements increase
Solution Approach 1:
The patent applies preliminary action by using the neural network to pre-process and transform complex data into representation data before clustering. This preliminary representation learning step prepares the data in advance, making it suitable for subsequent Gaussian mixture model clustering and improving overall accuracy
Solution Approach 2:
The patent achieves universality by creating a hybrid system that combines the capabilities of neural networks for feature extraction with Gaussian mixture models for clustering. This multi-functional approach leverages the strengths of both methods: the neural network handles complex non-Gaussian data structures while the GMM provides probabilistic clustering with automatic determination of cluster number
Data Source
AI summary
A learning method, executed by a computer including a memory and a processor, includes: inputting a plurality of items of data, and a plurality of labels representing clusters to which the plurality of items of data belong; converting each of the plurality of items of data by a predetermined neural network, to generate a plurality of items of representation data; clustering the plurality of items of representation data; calculating a predetermined evaluation scale indicating performance of the clustering, based on the clustering result and the plurality of labels; and learning a parameter of the neural network, based on the evaluation scale.