Machine Learning Model Selection for Small-Data Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning methods face challenges in generating trained models with good performance, particularly due to the difficulty in adjusting hyperparameters and the inefficiency of collecting large training data sets, which can lead to decreased production efficiency in image recognition tasks.
Innovation Solution
A machine learning device and method that acquire and evaluate training data, generate multiple trained models using various hyperparameter sets, select the best models based on performance, and perform inference calculations to generate and combine inference result candidates, thereby reducing the need for extensive data collection and improving model performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large quantity of training data is collected to improve model performance, then the performance of the trained model is improved, but the time and effort required for data collection increases significantly
Solution Approach 1:
The system performs preliminary actions by generating multiple candidate trained models with different hyperparameter sets before the actual inference task. This allows the selection of the most suitable model without requiring extensive training data collection, as the diversity of hyperparameter configurations compensates for limited data quantity.
Solution Approach 2:
The invention changes parameters by systematically varying hyperparameter sets across multiple training iterations. Instead of relying on large datasets, the system explores different parameter configurations (learning rates, batch sizes, network architectures) to find optimal models that perform well with smaller training data quantities.
2Reliability
If various training parameters (hyper parameters) are adjusted to improve model performance, then the performance may be improved, but the time spent on adjustment increases and good performance is not always obtained
Solution Approach 1:
The hyperparameter adjustment process is segmented into multiple independent training jobs, each with a specific hyperparameter set. Instead of manually adjusting parameters iteratively, the system divides the search space into discrete configurations and trains multiple models in parallel, making the parameter exploration systematic and efficient.
Solution Approach 2:
The system performs self-service by automatically generating and evaluating multiple hyperparameter configurations without requiring manual intervention. The automated evaluation and selection process eliminates the need for expert tuning, reducing both the time and expertise required for hyperparameter adjustment while consistently producing high-performance models.
3Adaptability or versatility
If ensemble training based on multiple trained models is performed to detect various objects, then the detection capability is improved, but the complexity of the training process increases
Solution Approach 1:
The system merges multiple trained models with different hyperparameter configurations into an ensemble system. By combining the strengths of various models trained on the same data with different parameters, the system achieves improved detection capability and robustness without requiring multiple separate training pipelines.
Solution Approach 2:
The training framework achieves universality by using a single training pipeline that can generate multiple models with different hyperparameter sets. This multi-functional approach allows the same training infrastructure to produce diverse models suitable for various detection tasks, reducing overall system complexity while maintaining versatility.
Data Source
AI summary
Good performance is obtained even with a small amount of training data, by reducing time and labor required for collecting training data for use in training. This machine learning device has an acquisition unit that acquires inference data and training data for use in machine learning, a training unit that performs machine learning based on the training data and sets of training parameters and generates trained models, a model assessment unit assesses whether or not the trained results by the trained models are good and displays the assessment results, a model selection unit that can receive a selected trained model, an inference calculation unit that performs inference calculation processing based on at least a part of the trained models and the inference data and generates inference result candidates, and an inference determination unit that outputs at least part of the inference result candidates or the combinations thereof.