Waveform Peak Picking With Explainable Learning Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing peak picking methods using machine learning lack explanatory properties for their derived results, necessitating a more accurate and illustrative approach.
Innovation Solution
A method for producing learning data that involves acquiring reference waveforms, specifying peak parts according to certain criteria, and training an estimation model using these data to output peak information in target waveforms, utilizing techniques like semantic segmentation and deep learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning techniques are used for peak picking, then accuracy of derived results is improved, but explanatory property of derived results deteriorates
Solution Approach 1:
The patent segments the peak picking task into multiple components: peak detection, peak classification, and result explanation. By dividing the complex machine learning process into separable stages, the system can maintain high accuracy through deep learning while preserving explanatory properties by separately handling the interpretation and visualization of results.
Solution Approach 2:
The patent introduces an intermediary explanation layer between the machine learning model and the final output. This intermediary component translates the black-box machine learning decisions into interpretable formats, such as visual indicators showing which features contributed to peak detection, thereby maintaining both accuracy and explanatory property.
2Loss of information
If conventional signal processing techniques are used for peak picking, then explanatory property of results is maintained, but accuracy of derived results deteriorates
Solution Approach 1:
The patent merges the advantages of conventional signal processing with machine learning techniques. By combining traditional methods that provide interpretability with modern deep learning that provides accuracy, the system achieves both high measurement precision and good explanatory property in peak picking results.
Solution Approach 2:
The patent creates a composite approach by integrating multiple methodologies: conventional signal processing algorithms are combined with machine learning models. This composite strategy allows the system to leverage the interpretability of traditional methods while achieving the superior accuracy of modern AI techniques.
3Productivity
If automated peak picking using machine learning is implemented, then productivity is improved, but device complexity deteriorates
Solution Approach 1:
The patent implements self-service mechanisms where the machine learning model automatically performs peak detection and classification without requiring complex manual configuration. The system learns from training data and autonomously adapts to different chromatogram types, reducing the operational complexity despite the advanced algorithms employed.
Solution Approach 2:
The patent performs preliminary actions by pre-training the machine learning model on extensive training data before actual peak picking operations. This preliminary training phase enables the model to handle diverse peak patterns automatically during operation, simplifying the real-time processing complexity while maintaining high productivity.
Data Source
AI summary
An analysis device produces learning data for training an estimation model. More specifically, the analysis device obtains a plurality of reference waveforms. In addition, the analysis device specifies information about a peak for each of the plurality of reference waveforms according to a certain criterion. The analysis device assigns the specified information about the peak to each of the plurality of reference waveforms.