Flow-Based Model Anomaly Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning systems face challenges in efficiently recognizing anomalies and continuing time series data, particularly in technical systems, due to limitations in predictive power and complexity management.
Innovation Solution
A computer-implemented method using a parameterizable generic flow-based model with a concatenation of submodules, where model parameters are learned progressively and fixed successively, allowing for improved predictive power and reduced complexity by distributing output data according to a predetermined probability distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a chain of submodules of predetermined length is trained so that the output exhibits a predetermined probability distribution, then the system can achieve the desired output distribution, but the device complexity and training difficulty increase significantly
Solution Approach 1:
The machine learning system is divided into multiple submodules (first submodule, second submodule, third submodule) that are trained successively in a specific flow direction. Each submodule processes data sequentially, with earlier submodules being trained and fixed before training subsequent submodules. This segmentation reduces the overall training complexity while maintaining the ability to achieve the desired predetermined probability distribution at the output.
2Device complexity
If model parameters are learned progressively with successive submodules being fixed, then the training complexity is reduced, but the training time and number of processing steps increase
Solution Approach 1:
The training process performs preliminary action by training and fixing the first submodule before proceeding to subsequent submodules. This preliminary training establishes a stable foundation that simplifies the training of later submodules, reducing the overall computational complexity and making the training process more manageable despite the multiple sequential steps required.
Data Source
AI summary
A computer-implemented method for training a machine learning system. The method includes: providing at least one training data set that includes a number of numerical vectors; propagating numerical values of the at least one training data set by a parameterizable generic flow-based model, the parameterizable generic flow-based model including a concatenation of at least two parameterizable submodules, each submodule being one parameterizable function each; and learning the model parameter of the parameterizable generic flow-based model; parameterizations of each parameterizable submodule being learned successively in the flow direction and being fixed before parameterizations of the parameterizable submodule next in the flow direction are learned, and the learning being directed at output data of each submodule being distributed according to a predetermined probability distribution.