Autoencoder Anomaly Detection via Genetic Algorithm Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost of hiring data scientists and the resource-intensive process of using advanced machine learning techniques to develop accurate neural networks for monitoring device anomalies make it challenging for organizations to efficiently detect anomalous operational states in devices.
Innovation Solution
The use of a genetic algorithm combined with selective optimization training to generate and train autoencoders, which enables faster and more resource-efficient detection of anomalous operations by iteratively evolving and training models to improve performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advanced machine learning techniques are used to develop accurate neural networks for anomaly detection, then detection accuracy is improved, but computing resource consumption increases significantly
Solution Approach 1:
The patent segments the neural network training process into two distinct phases: (1) a genetic algorithm phase that evolves network architecture and parameters without heavy computation, and (2) a selective backpropagation phase that trains only the most promising networks. This segmentation reduces overall computing resource consumption while maintaining detection accuracy.
Solution Approach 2:
The genetic algorithm performs preliminary actions by pre-evolving and selecting optimal network architectures and parameters before the computationally intensive backpropagation training. This preliminary selection ensures that only the most promising networks undergo full training, significantly reducing total computing resource requirements.
2Reliability
If traditional backpropagation training is used alone, then model training can be performed, but convergence to acceptable neural networks requires a large number of iterations and significant computing resources
Solution Approach 1:
The genetic algorithm performs preliminary evolution of network architectures and parameters before backpropagation training, pre-selecting promising candidates. This preliminary action reduces the number of iterations needed during the actual training phase, decreasing training time while maintaining effectiveness.
Solution Approach 2:
The system implements feedback by evaluating network performance after genetic evolution and using this information to guide selective backpropagation training. Only networks that show promise after genetic evolution undergo full training, creating an efficient feedback loop that reduces unnecessary computing iterations.
3Measurement precision
If data scientists manually interpret data sets and construct models, then model accuracy can be optimized, but hiring data scientists is cost prohibitive for many organizations
Solution Approach 1:
The system implements self-service by automatically performing the tasks traditionally requiring data scientists (model architecture design, parameter optimization, and selection) through the genetic algorithm framework. This automation eliminates the need to hire expensive data scientists while maintaining model accuracy through algorithmic optimization.
Solution Approach 2:
The patent substitutes the mechanical system of human data scientists with an automated genetic algorithm system. The genetic algorithm performs model construction and optimization tasks that previously required human expertise, replacing manual intervention with automated computational processes that are more cost-effective.
Data Source
AI summary
A method includes obtaining sensor data associated with operation of one or more devices and providing input data based on the sensor data to a dimensional-reduction model having an encoder portion and a decoder portion and configured such that the encoder portion is not mirrored by the decoder portion. The method also includes obtaining output data from the dimensional-reduction model responsive to the input data and determining a reconstruction error indicating a difference between the input data and the output data. The method also includes performing a comparison of the reconstruction error to an anomaly detection criterion and generating an anomaly detection output for the one or more devices based on a result of the comparison.