Probabilistic Model for Geophysical Data Feature Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing geophysical data to identify target features, such as marker shale bands in mines, are labor-intensive and prone to errors, and the use of machine learning techniques like Gaussian processes is inefficient due to the slow process of preparing training libraries and potential inaccuracies in predictions.
Innovation Solution
A computer-implemented method generates a probabilistic model using a training library to identify target features in geophysical data sets, allowing for the application of Gaussian processes and an active learning approach to improve the selection and modification of training examples, enhancing the accuracy and efficiency of feature identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection of geophysical data is performed, then accuracy of target feature identification can be maintained through expert judgment, but labor intensity and time consumption increase significantly
Solution Approach 1:
The patent replaces manual mechanical inspection by geologists with an automated computational system using machine learning algorithms. The system processes geophysical data through trained models that automatically identify target features, eliminating the need for manual visual inspection while maintaining consistent accuracy standards.
Solution Approach 2:
The system enables self-service by allowing the geophysical data processing to perform its own analysis through automated machine learning models. The trained models independently identify target features without requiring continuous human intervention, making the system autonomous in its core function.
2Productivity
If machine learning techniques like Gaussian processes are used to automate detection, then processing efficiency improves, but the training process becomes slow and complex
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models with comprehensive training libraries before actual data processing. This upfront preparation creates ready-to-use models that can quickly process new geophysical data without requiring time-consuming training during operational phases.
Solution Approach 2:
The system optimizes training efficiency by adjusting parameters such as training data selection criteria, model architecture configurations, and hyperparameter settings. These parameter changes reduce the computational burden and time required for training while maintaining model accuracy.
3Reliability
If comprehensive training libraries are created to improve model accuracy, then prediction reliability increases, but the complexity and time required for model preparation increases
Solution Approach 1:
The patent segments the training library creation process into manageable components: data collection, data labeling, feature extraction, model training, and validation. This segmentation allows each step to be optimized independently and reduces overall complexity by breaking down the monolithic training process into discrete, controllable tasks.
Solution Approach 2:
The system introduces intermediary components such as data preprocessing pipelines, feature engineering modules, and cross-validation mechanisms that mediate between raw data and final model training. These intermediaries simplify the overall process by handling complex transformations and validations automatically.
Data Source
AI summary
Described herein is a computer implemented method for generating a probabilistic model usable to identify instances of a target feature in geophysical data sets stored on a memory device. The computer implemented method using a computer processing unit to generate a probabilistic model from a training library for use in identifying instances of the target feature in the geophysical data sets, applying, using the computer processing unit, the probabilistic model to one or more of the geophysical data sets to generate a plurality of results, processing, using the computer processing unit, the set of results according to an acceptability criteria in order to identify a plurality of candidate results, receiving a selection of one or more of the candidate results and for the or each selected candidate result displaying on a display the result and its associated geophysical data set to assist a user in making an assessment as to whether or not the probabilistic model is an acceptable model for the processing of the geophysical data sets, receiving from a user an assessment as to whether or not the probabilistic model is an acceptable model; and if the assessment received indicates the probabilistic model is an acceptable model for processing the geophysical data, outputting the probabilistic model and/or the training library.