Roundtrip: a deep generative neural density estimator

Datasets used in Roundtrip paper. (https://arxiv.org/abs/2004.09017) UCI datasets: AReM/data.npy: Activity Recognition data collected from Multi sensors. Data size (42239, 6) CASP/data.npy: Protein Tertiary Structure data. Data size (45730, 9) BANK/data.npy: Financial data collected from clients. Data size (45211, 17) YPMSD/data.npy: Audio data of various songs. Data size (515345,90) HEPMASS/train.csv and HEPMASS/test.csv: Data of high-energy particles. OOD contains three outlier detection datasets Shuttle, Mammography and ForestCover.

《Roundtrip》论文所使用的数据集（论文链接：https://arxiv.org/abs/2004.09017）：UCI数据集包含AReM/data.npy、CASP/data.npy、BANK/data.npy及YPMSD/data.npy；其中AReM/data.npy为多传感器采集的活动识别数据，数据尺寸为(42239, 6)，CASP/data.npy为蛋白质三级结构数据，数据尺寸为(45730, 9)，BANK/data.npy为从客户处采集的金融数据，数据尺寸为(45211, 17)，YPMSD/data.npy为各类歌曲的音频数据，数据尺寸为(515345, 90)。HEPMASS/train.csv与HEPMASS/test.csv为高能粒子数据；OOD（分布外，Out-of-Distribution）包含3个离群检测数据集，分别为Shuttle、Mammography及ForestCover。