Pathwise Derivatives Beyond the Reparameterization Trick

Authors: Martin Jankowiak, Fritz Obermeyer

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate with a variety of synthetic experiments and stochastic variational inference tasks that our pathwise gradients are competitive with other methods.
Researcher Affiliation Industry 1Uber AI Labs, San Francisco, USA. Correspondence to: <jankowiak@uber.com>, <fritzo@uber.com>.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Our approximations for pathwise gradients for the Gamma, Beta, and Dirichlet distributions are available in the 0.4 release of Py Torch (Paszke et al., 2017).
Open Datasets Yes The dataset we consider is the Olivetti faces dataset,16 which consists of 64 64 grayscale images of human faces. 16http://www.cl.cam.ac.uk/research/dtg/ attarchive/facedatabase.html
Dataset Splits No The paper describes datasets but does not provide specific information about training, validation, or test splits. No explicit mention of a validation set or its size is made.
Hardware Specification No The paper does not provide any specific hardware details such as CPU, GPU models, or memory specifications used for running its experiments.
Software Dependencies Yes Our approximations for pathwise gradients for the Gamma, Beta, and Dirichlet distributions are available in the 0.4 release of Py Torch (Paszke et al., 2017).
Experiment Setup No The paper provides some model-specific parameters but lacks concrete experimental setup details such as learning rates, batch sizes, number of epochs, or optimizer settings in the main text.