Exemplar Vector Selection for ML Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning (ML) models used for anomaly detection in time series readings face a challenge in balancing prognostic accuracy and compute costs, as reducing the number of training vectors to reduce costs also decreases accuracy, making it difficult to determine the optimal number of exemplar vectors for training without incurring excessive resource constraints.
Innovation Solution
An inferential exemplar selection system automatically determines the quantity of exemplar vectors based on relationships between training data and ML model performance, selecting a balance that maximizes accuracy while minimizing compute resources, using a combination of training vector identifiers, boost function selectors, selection quantity generators, and ML model trainers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If the number of training vectors is reduced to decrease compute costs, then compute resources are saved, but prognostic accuracy of the ML model decreases
Solution Approach 1:
The system automatically determines the optimal quantity of exemplar vectors through self-service mechanisms. The ML model trainer autonomously selects the quantity of exemplar vectors based on the available training vectors, eliminating the need for manual tuning while balancing compute cost and accuracy requirements
Solution Approach 2:
The system changes the parameter of exemplar vector quantity dynamically. By adjusting the quantity of exemplar vectors based on available training data and performance requirements, the system optimizes the balance between compute resource consumption and model accuracy
2Measurement precision
If the number of exemplar vectors is increased to improve prognostic accuracy, then model accuracy improves, but compute resources required for training and operation increase
Solution Approach 1:
The system applies partial action by selecting only the necessary quantity of exemplar vectors rather than using all available training vectors. This partial selection is sufficient to achieve the required prognostic accuracy while avoiding the excessive compute costs associated with using more vectors than necessary
3Adaptability or versatility
If manual determination of exemplar vector quantity is used, then flexibility in tuning is achieved, but system complexity and resource constraints increase
Solution Approach 1:
The system replaces manual determination with self-service automation. The ML model trainer automatically determines the optimal quantity of exemplar vectors based on the available training vectors, maintaining adaptability while reducing system complexity and eliminating manual intervention requirements
Data Source
AI summary
Systems, methods, and other embodiments associated with determining a quantity of exemplar vectors to select from available training vectors are described. In one embodiment, a method includes determining an available quantity of training vectors that are available in a set of time series signals. A boost function is automatically selected from a plurality of different boost functions based on the available quantity of the training vectors. A selection quantity of the exemplar vectors to select from the training vectors is generated by applying the selected boost function to the training vectors. A quantity of the exemplar vectors is selected from the training vectors based on the selection quantity. A machine learning model is trained to detect an anomaly in the time series signals based on the exemplar vectors that were selected.