Deep Learning Architecture Search with Automated Data Pre-processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems, particularly deep learning systems, face bottlenecks in data pre-processing and architecture search, as they require large amounts of clean data and manual intervention, and existing techniques do not account for data quality or tailor pre-processing strategies to specific architectures.
Innovation Solution
A combined data pre-processing and architecture search method that uses a multi-objective learning process to select an optimal deep learning architecture and data pre-processing strategy simultaneously, improving data quality and efficiency through automated end-to-end systems that include a multi-objective learning module for extracting properties and ranking architectures based on task and data characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual data pre-processing is performed to improve data quality, then data quality improves, but time consumption and labor costs increase
Solution Approach 1:
The system performs automated data pre-processing by having the selected deep learning architecture automatically clean and transform the data according to the selected pre-processing strategy, eliminating the need for manual human intervention in data preparation tasks
Solution Approach 2:
The system changes the state of data from raw/unprocessed to cleaned/processed by applying selected pre-processing operations (such as handling missing values, removing stop words, correcting errors) automatically through the multi-objective learning framework
2Reliability
If multiple data pre-processing strategies are evaluated to find the optimal one, then data quality improves, but system complexity increases
Solution Approach 1:
The system merges the data pre-processing strategy selection with the deep learning architecture selection into a single multi-objective learning framework, allowing both to be optimized simultaneously rather than separately, thereby managing complexity through integration
Solution Approach 2:
The multi-objective learning framework serves multiple functions: it selects both the appropriate deep learning architecture and the suitable data pre-processing strategy, making the system versatile in handling different data types and tasks through a unified approach
3Ease of operation
If existing architecture search techniques are used that assume clean input data, then architecture search is simplified, but data quality issues are ignored
Solution Approach 1:
The system performs preliminary data pre-processing actions automatically by selecting and applying appropriate pre-processing strategies before the deep learning architecture processes the data, ensuring data quality issues are addressed in advance rather than assumed to be already resolved
Data Source
AI summary
Methods, systems, and computer program products for combined data pre-process and architecture search for deep learning models are provided herein. A computer-implemented method includes obtaining data corresponding to a deep learning task; performing, based on the deep learning task and the data, a multi-objective learning process to select an optimal combination of (i) a deep learning architecture for the deep learning task and (ii) a data pre-processing strategy to be applied to the data, the data pre-processing strategy comprising one or more pre-processing steps; pre-processing the data for the selected deep learning architecture based on the data pre-processing strategy; and providing the pre-processed data as input to the selected deep learning architecture to perform the deep learning task.