Scholarlink

This study introduces a custom-designed Convolutional Neural Network (CNN) aimed at enhancing classification performance while minimizing computational overhead for image recognition tasks. The proposed model integrates advanced techniques such as batch normalization, max-pooling, and dynamic dropout layers to address overfitting and stabilize training. This approach explores the impact of various hyperparameters, including learning rates, weight initialization strategies, and optimizers, to systematically optimize the training process. Comprehensive experiments conducted on the CIFAR-10 highlights dataset validate the efficacy of the proposed custom architecture in achieving superior accuracy and efficient feature extraction compared to standard CNN models. The study highlights the role of hyperparameter tuning in optimizing deep learning models for real-world applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimized CNN Architecture for Enhanced Tiny Image Classification Using Hyperparameter Tuning and Regularization

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Aluri Jitendra Chowdary,
  • Mukesh Medikonda,
  • Krishaiah Narukulla,
  • Jyostna Devi Bodapati

摘要

This study introduces a custom-designed Convolutional Neural Network (CNN) aimed at enhancing classification performance while minimizing computational overhead for image recognition tasks. The proposed model integrates advanced techniques such as batch normalization, max-pooling, and dynamic dropout layers to address overfitting and stabilize training. This approach explores the impact of various hyperparameters, including learning rates, weight initialization strategies, and optimizers, to systematically optimize the training process. Comprehensive experiments conducted on the CIFAR-10 highlights dataset validate the efficacy of the proposed custom architecture in achieving superior accuracy and efficient feature extraction compared to standard CNN models. The study highlights the role of hyperparameter tuning in optimizing deep learning models for real-world applications.