Neural Network Data Adjustment System for Learning Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning technologies fail to ensure that data used for learning is both suitable and effective, as they do not adequately assess whether data is appropriate for learning, leading to models like neural networks not achieving desired performance due to the inclusion of unsuitable data.
Innovation Solution
A data adjustment system that measures the influence of learning data on a neural network, excludes data with low influence, and acquires or generates new data with high influence to optimize the data set used for learning, thereby ensuring that only relevant data contributes to model improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional machine learning uses all available data including data with missing values, then the data quantity is sufficient, but the model performance deteriorates due to inclusion of unsuitable data
Solution Approach 1:
The patent extracts and removes unsuitable data (data with missing values or low learning contribution) from the training dataset. The measurement unit identifies data with missing values, and the adjustment unit excludes this data to improve model performance while maintaining sufficient data quantity for effective learning.
Solution Approach 2:
The patent changes the parameter of data quality by measuring the degree of influence of each data point on learning. The adjustment unit modifies the dataset by adding data with high influence and removing data with low influence, thereby transforming the dataset from a simple collection to an optimized learning resource.
2Loss of information
If data with missing values is complemented with candidate values, then the data completeness is improved, but the data quality deteriorates due to potential introduction of inappropriate values
Solution Approach 1:
Instead of attempting to fill missing values (which could introduce errors), the patent converts the harmful effect of missing values into a beneficial filtering mechanism. Data with missing values are identified as having low learning contribution and are excluded from the training set, transforming a data quality problem into a data selection opportunity.
3Ease of operation
If all data is used for learning without selection, then the learning process is simple, but the learning efficiency deteriorates due to inclusion of irrelevant data
Solution Approach 1:
The patent introduces a feedback mechanism where the measurement unit evaluates the degree of influence of each data point on learning outcomes. This feedback information is used by the adjustment unit to optimize the dataset, creating a closed-loop system that continuously improves learning efficiency based on measured performance impact.
Solution Approach 2:
The system performs self-optimization by automatically measuring data influence and adjusting the dataset without requiring manual intervention. The measurement unit and adjustment unit work autonomously to identify and exclude low-value data, making the learning process more efficient while maintaining operational simplicity.
Data Source
AI summary
A data adjustment system according to the present disclosure includes an information processing apparatus, and a terminal device, in which the information processing apparatus includes, a measuring unit configured to measure a degree of influence of learning data on learning in a neural network, the learning data being used for the learning, and an adjustment unit configured to adjust the learning data by excluding data measured as having a low degree of influence, acquiring new data from the terminal device or a database, or adding the acquired new data, the new data being data to be newly added corresponding to data measured as having a high degree of influence.