Transfer Learning Data Processing for Small Dataset Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data processing technologies face challenges in achieving highly accurate data processing, particularly when dealing with small-scale data sets that lack sufficient information, as they often rely solely on target data without leveraging larger datasets with abundant information.
Innovation Solution
A data processing device and method that employs transfer learning to derive evaluation indices by combining small-scale and large-scale data sets, using synthetic machine learning models to enhance accuracy, where the processor acquires and processes data to generate evaluation indices based on regression labels and synthetic regression labels derived from both sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data processing relies solely on small-scale target data, then the system is simple and easy to operate, but the processing accuracy is insufficient due to lack of information
Solution Approach 1:
The patent introduces a transfer learning mechanism that acts as an intermediary between source data and target data. The system uses a processor to perform transfer learning operations, combining source data with target data to generate improved regression labels. This intermediary process enables the system to leverage information from large-scale source data while maintaining focus on small-scale target data, thereby improving accuracy without requiring complete redesign of the processing system.
Solution Approach 2:
The patent merges source data and target data through transfer learning operations. The processor combines information from both data sources to generate synthetic regression labels that incorporate insights from both small-scale target data and large-scale source data. This merging approach allows the system to achieve high processing accuracy by synthesizing information from both data types while maintaining a relatively simple overall system structure.
2Measurement precision
If only target data is used for processing, then data collection is simple, but processing accuracy is insufficient due to inadequate information volume
Solution Approach 1:
The transfer learning processor serves as an intermediary that bridges the gap between limited target data and the need for sufficient information. It processes both source data and target data through learned transformations to generate enhanced regression labels. This intermediary mechanism enables the system to achieve high regression label accuracy by effectively utilizing information from both data sources, overcoming the limitation of small data quantity.
Solution Approach 2:
The system performs preliminary transfer learning operations to pre-process and enhance the information content before final regression label generation. By conducting these preliminary actions on the combined source and target data, the system prepares enriched regression labels that contain sufficient information for accurate processing, thereby improving accuracy without requiring a large volume of raw data.
3Measurement precision
If transfer learning with synthetic models is applied, then processing accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies transfer learning and synthetic model generation selectively rather than universally. The processor performs these computationally intensive operations only when and where needed to enhance regression label accuracy. By applying these advanced techniques partially rather than completely, the system improves processing accuracy in critical areas while minimizing the overall time cost associated with complex computational operations.
Data Source
AI summary
According to one embodiment, a data processing device includes an acquisitor, and a processor. The acquisitor is configured to acquire a first acquired data and a first other data. The processor is configured to perform a first evaluation index derivation operation deriving a first evaluation index from a plurality of first regression labels and a plurality of first synthetic regression labels. The first regression labels are derived from a plurality of first machine learning models. The first machine learning models are derived from a plurality of first sample data.The first synthetic regression labels are derived from a plurality of first synthetic machine learning models. The first synthetic machine learning models are derived from the first sample data and the first acquire data by a first transfer leaning. The first sample data are derived from the first other data or a first conversion other data.