Multi-Sampling Model Training for Robust Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limited quantity of data samples for the 'Ant Check Later' anti-cash out model results in ineffective model training due to insufficient validation, leading to reduced prediction accuracy.
Innovation Solution
The multi-sampling model training method involves performing repeated random sampling to generate multiple training and validation sets, training models on each set, evaluating their performance, retaining models that meet predetermined criteria, and combining their predictions to produce a more robust output model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the proportion of the training set in the sample data set is increased, then the model training is improved, but the size of the validation set is reduced and the model effectiveness validation is insufficient
Solution Approach 1:
The patent applies segmentation by dividing the limited sample data into multiple subsets through multi-sampling techniques. Instead of a single train-validation split, the data is segmented into multiple training sets and validation sets across different sampling iterations. This allows the model to be trained on multiple segments while maintaining multiple validation segments for comprehensive evaluation, resolving the contradiction between training set size and validation set size.
Solution Approach 2:
The patent implements periodic action through repeated sampling and model training cycles. Multiple models are trained periodically on different sampled data subsets, and their performances are evaluated on corresponding validation sets. This periodic multi-cycle approach allows sufficient model training while maintaining adequate validation capacity across different iterations, addressing the contradiction between training effectiveness and validation sufficiency.
2Quantity of substance
If the proportion of the training set in the sample data set is decreased, then the validation set size is increased, but the model training is insufficient and prediction accuracy is reduced
Solution Approach 1:
The patent segments the limited data into multiple subsets across different sampling iterations. Each iteration uses a portion of the data for training while reserving validation sets, but the cumulative effect of multiple iterations provides both sufficient training data aggregate and adequate validation data. This segmentation approach ensures that model prediction accuracy is improved through multiple training exposures while validation set size remains sufficient across iterations.
Solution Approach 2:
Through periodic repeated sampling and training cycles, the model accumulates training experience across multiple iterations. Each periodic cycle uses a manageable training set size, but the aggregate training across all cycles provides sufficient training data. Simultaneously, validation sets are maintained in each cycle to ensure adequate validation capability, resolving the contradiction between training sufficiency and validation size.
3Reliability
If multi-sampling and combined model training are performed, then the robustness and stability of the model are improved, but the training time and computational resources are increased
Solution Approach 1:
The patent applies partial action by performing multi-sampling with a predetermined number of iterations rather than exhaustive sampling. Models are trained on multiple sampled subsets, and combined model training is performed on the top-performing models based on validation results. This partial approach achieves sufficient model robustness without the excessive computational burden of trying all possible samples, balancing reliability improvement with time efficiency.
Solution Approach 2:
The patent implements feedback mechanisms where models are evaluated on validation sets after each training iteration, and performance results feed into the selection process for combined model training. Models that meet predetermined elimination criteria are retained for combination, while poor-performing models are eliminated. This feedback loop ensures that computational resources are focused on promising models, improving robustness efficiently without wasting time on inadequate models.
Data Source
AI summary
The present disclosure provides multi-sampling model training methods and devices. One exemplary training method includes: performing multi-sampling on samples to obtain a training set and a validation set in each sampling; using the training set and the validation set obtained in each sampling as a group, and performing model training and obtaining a trained model using the training set in each group; evaluating the trained model using the training set and the validation set in each group separately; eliminating or retaining the trained model based on the evaluation results and a predetermined elimination criterion; obtaining prediction results of the samples using retained models; and obtaining a final model by performing combined model training on the retained models using the prediction results. The final model obtained using embodiments of the present disclosure can be more robust and stable, and can provide more accurate prediction results, thus greatly improving efficiency of modeling.