Estimation Model for Parameter Value Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques face difficulties in easily determining the parameter value at which a target event occurs, such as a drug reaction, due to the complexity in correlating image analysis with the occurrence of the event.
Innovation Solution
A system comprising a learning apparatus that generates learning data sets with attribute values based on image thresholds, performs deep learning processing on estimation models, and determines the occurrence parameter value by evaluating the estimation error, allowing for precise determination of the parameter value at which a target event occurs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image analysis techniques are used to determine parameter values at which target events occur, then image processing can be performed, but the determination of the parameter value becomes complex and difficult
Solution Approach 1:
The patent introduces an estimation model as an intermediary between image analysis and parameter determination. The model learns the relationship between image features and parameter values through training, then automatically estimates parameter values from new images without requiring complex manual analysis. This intermediary model simplifies the overall system while improving measurement precision.
Solution Approach 2:
The patent performs preliminary learning processing to train the estimation model before actual use. During this preliminary phase, the model learns the relationship between images and parameter values from training data. Once trained, the model can quickly and accurately determine parameter values without requiring complex analysis during operation, thus reducing operational complexity while maintaining high precision.
2Reliability
If conventional image analysis is performed to correlate images with event occurrence, then image processing can be done, but the ease of determining parameter values deteriorates
Solution Approach 1:
The estimation model performs self-service by automatically learning the relationship between images and parameter values during training, then independently determining parameter values for new images without requiring complex manual intervention. This self-service capability improves ease of operation while maintaining reliable drug reaction analysis through the model's learned patterns.
Solution Approach 2:
The patent transforms the complex task of parameter determination into a parameter estimation problem that can be solved by changing the approach from direct analysis to model-based prediction. By adjusting the model's internal parameters during training and then using these learned parameters for estimation, the system achieves both reliability and ease of operation.
3Measurement precision
If deep learning processing is performed on multiple estimation models, then the accuracy of parameter determination improves, but the processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by performing deep learning processing selectively - training multiple estimation models during the preliminary phase, then using the trained models for rapid inference. The excessive action体现在 training multiple models to ensure accuracy, but once trained, the models provide fast predictions without requiring repeated extensive processing for each new image.
Solution Approach 2:
The patent performs all computationally intensive deep learning processing in advance during the training phase. Once the estimation models are trained with high accuracy, they can quickly determine parameter values for new images without requiring repeated heavy processing. This preliminary action shifts the computational burden to an initial phase, reducing time loss during actual operation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In conventional technology, it takes time to determine a parameter value at which a target event occurs. An apparatus is provided, which includes an obtainment portion configured to obtain a plurality of images corresponding to respective values of a parameter that affects occurrence of a target event; a generation portion configured to generate, for each of a plurality of thresholds relating to the parameter, a learning data set in which at least a part of the plurality of images is each given an attribute value, the attribute value varying based on whether the threshold is exceeded; a learning processing portion configured to perform, for each generated learning data set, learning processing on an estimation model that estimates an attribute of an image in response to input of the image; and an evaluation value output portion configured to output, for each estimation model, an evaluation value according to an estimation error of the estimation model.