Summary of hyperparameter optimization methods used in MCount. The table includes the method number, image type (labeled or unlabeled), number of images used for optimization, hyperparameter <i>d</i> optimization method, and hyperparameter <i>λ</i> optimization method. Method 1 uses grid search with cross-validation on 960 labeled images (Section 3.2 & 3.3). Method 2 and method 3 decoupled the optimization of <i>d</i> and <i>λ</i> and use the same empirical method for determining <i>d</i> (Section 3.3). Method 2 uses the average <i>λ</i> of 10 ~ 20 labeled images (Section 3.3.1), while method 3 chooses the value of <i>λ</i> that leads to equidispersion on 40 ~ 50 unlabeled images (Section 3.3.2).
收藏NIAID Data Ecosystem2026-05-02 收录
数据链接：
官方服务：
资源简介：
Summary of hyperparameter optimization methods used in MCount. The table includes the method number, image type (labeled or unlabeled), number of images used for optimization, hyperparameter d optimization method, and hyperparameter λ optimization method. Method 1 uses grid search with cross-validation on 960 labeled images (Section 3.2 & 3.3). Method 2 and method 3 decoupled the optimization of d and λ and use the same empirical method for determining d (Section 3.3). Method 2 uses the average λ of 10 ~ 20 labeled images (Section 3.3.1), while method 3 chooses the value of λ that leads to equidispersion on 40 ~ 50 unlabeled images (Section 3.3.2).
创建时间：
2025-03-19
