Deep Learning Architecture Optimization Using Reusable Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning architectures face challenges in effectively classifying data objects, especially when the training data set is small, leading to inefficient and costly maintenance, particularly in complex AI tasks such as scene or language understanding.
Innovation Solution
A method and system for determining an optimized deep learning architecture using a small set of training data, which includes receiving training images and learning parameters to generate a deep learning model with reusable filters, enabling efficient health condition determination in medical diagnostics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning architectures use sufficient filters and layers for effective modeling, then classification accuracy is improved, but the amount of training data required increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-training deep learning models on large datasets to learn reusable filters and features before fine-tuning on small domain-specific datasets. This allows the model to start with pre-learned feature extraction capabilities, reducing the amount of domain-specific training data needed while maintaining high classification accuracy.
Solution Approach 2:
The patent implements universality by designing deep learning models with reusable filters and transferable features that can be applied across multiple tasks and domains. The same filter set learned for general image processing can be reused for specific medical imaging tasks, reducing the need for task-specific training data while maintaining effectiveness.
2Device complexity
If manual design of classification systems is performed, then system complexity is reduced, but effectiveness for large and complex data sets deteriorates
Solution Approach 1:
The patent applies self-service by enabling deep learning models to automatically learn optimal filter configurations and feature extraction parameters from data, rather than requiring manual design. The model self-adjusts its architecture and parameters during training, achieving high effectiveness on complex datasets while the complexity is managed through automated hyperparameter optimization and architecture search.
3Adaptability or versatility
If deep learning models are trained with sufficient filters and layers, then modeling capability is improved, but computational complexity and maintenance cost increase
Solution Approach 1:
The patent applies extraction by separating the feature extraction function from the classification function. Pre-trained filters are extracted and reused across multiple classification tasks, allowing the model to maintain high adaptability while reducing computational complexity by avoiding redundant feature learning in each new task.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting model architecture parameters such as filter counts, layer depths, and learning rates based on the specific task requirements and available data. This allows the model to maintain high modeling capability while adapting computational complexity to match the actual needs of each application, reducing unnecessary computational overhead.
Data Source
AI summary
A method for determining optimized deep learning architecture includes receiving a plurality of training images and a plurality of real time images corresponding to a subject. The method further includes receiving, by a medical practitioner, a plurality of learning parameters comprising a plurality of filter classes and a plurality of architecture parameters. The method also includes determining a deep learning model based on the plurality of learning parameters and the plurality of training images, wherein the deep learning model comprises a plurality of reusable filters. The method further includes determining a health condition of the subject based on the plurality of real time images and the deep learning model. The method also includes providing the health condition of the subject to the medical practitioner.