Research data supporting "Unveil the unseen: exploit information hidden in noise"

The datasets are training sets (file names ending with _train.csv) and validation sets (file names ending with _validate.csv). The machine learning algorithm is trained on the data in a _train.csv file and validated against the data in the corresponding _validate.csv file. The first column of each file is the input variable. The second column is the available values for the intermediate target variable. The last column of each file is the values of the final target variable.

The extrapolation_using_Y training and validation sets are used to validate the ability of the machine learning algorithm to perform extrapolation using intermediate target variable. The sinosoidal_noise training and validation sets are used to validate the ability of the machine learning algorithm to perform extrapolation using uncertainty in the intermediate target variable. The phase_transition and droplet_diffraction training and validation files contain the experimental data for the real-world physical examples that we applied the methodology to.

本数据集包括训练集（以_train.csv结尾的文件名）和验证集（以_validate.csv结尾的文件名）。机器学习算法在_train.csv文件中的数据上进行训练，并通过对应的_validate.csv文件中的数据进行验证。每个文件的第一个列是输入变量。第二个列是中间目标变量的可能值。每个文件的最后一列是最终目标变量的值。利用_Y训练集和验证集的外推能力验证了机器学习算法在利用中间目标变量进行外推的能力。正弦噪声训练集和验证集用于验证机器学习算法在利用中间目标变量不确定性进行外推的能力。相变和液滴衍射的训练集和验证文件包含了我们应用该方法的实际物理例子的实验数据。