reproducibilityindex.ai

On the Generative Utility of Cyclic Conditionals

Authors: Chang Liu, Haoyue Tang, Tao Qin, Jintao Wang, Tie-Yan Liu

Reproducibility Variable	Result	LLM Response
Research Type	Experimental	With the prior constraint removed, Cy Gen better ﬁts data and captures more representative features, supported by both synthetic and real-world experiments.
Researcher Affiliation	Collaboration	1 Microsoft Research Asia, Beijing, 100080. 2 Tsinghua University, Beijing, 100084.
Pseudocode	No	The paper describes methods using mathematical formulas but does not provide pseudocode or algorithm blocks.
Open Source Code	Yes	Codes: https://github.com/changliu00/cygen.
Open Datasets	Yes	We test the performance of Cy Gen on real-world image datasets MNIST and SVHN.
Dataset Splits	No	The paper mentions using MNIST and SVHN datasets but does not explicitly provide details about training, validation, or test splits.
Hardware Specification	No	The paper does not provide specific details about the hardware used to run the experiments.
Software Dependencies	No	The paper mentions software like PyTorch in its references but does not explicitly list the software dependencies with specific version numbers used for its implementation.
Experiment Setup	Yes	All models are trained by Adam optimizer [43] with a learning rate 10 3. ... For MNIST, the dimension of latent space d Z = 10; for SVHN, d Z = 32. For both datasets, the number of Sylvester ﬂows is 8, which means 8 Householder transformations. ... We train Cy Gen for 30,000 iterations for MNIST, and 100,000 for SVHN. Batch size is 128 for both datasets.