Backpropagation Process for Geophysical Data Long-Range Dependencies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional backpropagation-enabled processes for geophysical data sets have a limited field of view, which restricts their ability to capture long-range dependencies and accurately interpret complex geological structures.
Innovation Solution
The method involves dependency-training and interdependency-training of backpropagation-enabled processes to compute spatial and inter-feature relationships, followed by fusion of conditional featurized input channels to produce a combined representation, which is then used to predict geologic features, geophysical properties, and hydrocarbon occurrences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional backpropagation-enabled processes are used with limited field of view, then computational resources are reduced, but long-range dependencies cannot be captured accurately
Solution Approach 1:
The patent segments the geophysical data set into multiple patches and processes them through multiple backpropagation-enabled processes in parallel. Each process handles a specific patch, and the results are combined to achieve comprehensive long-range dependency capture across the entire data set, resolving the contradiction between processing complexity and dependency capture accuracy.
Solution Approach 2:
The patent introduces a new dimension by processing multiple patches simultaneously across different spatial locations and combining their outputs. This dimensional expansion allows the system to capture long-range dependencies without requiring a single monolithic process, thus reducing overall process complexity while improving measurement precision.
2Area of stationary object
If larger filter sizes are used to expand field of view, then long-range dependencies are captured better, but computational cost increases significantly
Solution Approach 1:
Instead of using a single large filter to cover the entire field of view, the patent segments the data into multiple patches and applies smaller filters to each patch independently. This segmentation approach achieves the same field of view coverage with significantly reduced computational cost for each individual processing operation.
Solution Approach 2:
The patent merges the outputs from multiple smaller filter operations across different patches to achieve the equivalent effect of a single large filter. This combining approach provides the desired field of view while maintaining lower computational costs associated with smaller individual filter operations.
3Area of stationary object
If recursively applying small filters is used to achieve wide field of view, then field of view is expanded, but computational time increases
Solution Approach 1:
The patent segments the geophysical data into multiple patches that can be processed in parallel. This segmentation allows simultaneous computation across different spatial regions, expanding the effective field of view while reducing total computational time compared to sequential recursive filter application.
Solution Approach 2:
The patent maintains continuous useful action by processing multiple patches concurrently through parallel execution of backpropagation-enabled processes. This continuous parallel processing achieves wide field of view coverage without the time penalty of sequential recursive operations.
Data Source
AI summary
A method for capturing long-range dependencies in geo-physical data sets involves dependency-training a first backpropagation-enabled process, followed by interdependency-training the dependency-trained backpropagation-enabled process. Dependency-training computes spatial relationships for each input channel of a geophysical data set. Interdependency-training computes inter-feature and spatial relationships between each of the featurized input channels. The output conditional featurized input channels are fused to produce a combined representation of the conditional featurized input channels. The combined representation is inputted to a second backpropagation-enabled process to compute a prediction selected from the group consisting of a geologic feature occurrence, a geophysical property occurrence, a hydrocarbon occurrence, an attribute of subsurface data, and combinations thereof.