IRIS Multiple Instance Learning Dataset

This dataset contains the data for the paper 'Using Multiple Instance Learning for Explainable Solar Flare Prediction' (arxiv pre-print) . It comes as a compressed Python Numpy-File and contains the following variables: Name Shape Description data (10'000, 1100, 240) 10'000 Bags of zero-padded spectrograms data_scaled (10'000, 1100, 240) Like data, but standard-scaled masks (10'000, 1100) Masks that indicate where spectrograms have been zero-padded groups (10'000,) Observation group the bag is assigned to obs_ids (10'000,) Observation ID the bag is assigned to obs_classes (10'000,) Observation class (AR/PF) the bag is assigned to raster_pos (10'000,) Raster position number the bag was taken from (always 0 for sit-and-stare) folds (10'000,) Validation fold for the particular observation group To load the e.g. the variable 'data', use Python and Numpy: <pre><code class="language-python">import numpy as np f = np.load("IRISMIL_dataset_10000_bags.npz", allow_pickle=True) f['data']</code></pre>本数据集配套于论文《使用多实例学习（Multiple Instance Learning）实现可解释性太阳耀斑预测》（arXiv预印本）。该数据集以压缩Python NumPy文件格式提供，包含以下变量： - `data`：形状为(10000, 1100, 240)，包含10000个零填充声谱图实例包。 - `data_scaled`：形状为(10000, 1100, 240)，与`data`格式一致，但经过了标准化缩放处理。 - `masks`：形状为(10000, 1100)，用于标记声谱图零填充位置的掩码矩阵。 - `groups`：形状为(10000,)，代表该实例包所属的观测分组。 - `obs_ids`：形状为(10000,)，代表该实例包对应的观测ID。 - `obs_classes`：形状为(10000,)，代表该实例包所属的观测类别（AR/PF）。 - `raster_pos`：形状为(10000,)，代表该实例包采集所在的光栅位置编号（定点凝视模式（sit-and-stare）下始终为0）。 - `folds`：形状为(10000,)，代表对应观测分组的验证折次。 若需加载如`data`这类变量，可使用Python与NumPy库执行以下代码： python import numpy as np f = np.load("IRISMIL_dataset_10000_bags.npz", allow_pickle=True) f['data']