Construction of Sports Training Performance Prediction Model Based on a Generative Adversarial Deep Neural Network...

Area of Science:

• Artificial Intelligence
• Computer Vision
• Machine Learning

Background:

• Generative adversarial networks (GANs) face challenges in image restoration, including gradient issues and training instability.
• Existing GAN-based image restoration methods often lack local consistency and require extensive training time.
• Predicting student performance necessitates robust feature extraction and analysis of behavioral data.

Purpose of the Study:

• To propose a multigenerative adversarial image restoration algorithm addressing GAN limitations.
• To develop an attention-based mechanism for enhanced student performance prediction.
• To improve the stability, efficiency, and quality of image restoration and performance prediction models.

Main Methods:

• A multigenerative adversarial image restoration algorithm utilizing multigranularity reconstruction sampling.
• Modifying generative network initialization and employing reconstruction sampling to stabilize gradients.
• Implementing segmentation invariance to reduce training time and introducing an algorithm adaptability metric.
• Extracting deep student behavioral features via a generative adversarial deep neural network and applying maximum pooling.
• Integrating a temporal attention mechanism for student performance prediction.

Main Results:

• Reconstruction sampling effectively stabilizes GAN training and gradients in image restoration.
• Segmentation invariance significantly shortens training duration without compromising restored image quality.
• The proposed algorithm adaptability metric provides a comprehensive evaluation of image restoration.
• The attention-based mechanism improves the selection of salient student behavioral features.
• The temporal attention mechanism enhances student performance prediction by weighting weekly behavioral data.

Conclusions:

• The novel multigenerative adversarial algorithm offers a stable and efficient solution for image restoration.
• The attention-based approach provides a more accurate and nuanced method for student performance prediction.
• This research contributes to advancements in both image processing and educational data mining.