Learning with Stochastic Orders

Authors: Carles Domingo-Enrich, Yair Schiff, Youssef Mroueh

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide a min-max framework for learning with stochastic orders and validate it experimentally on synthetic and high-dimensional image generation, with promising results.
Researcher Affiliation Collaboration Carles Domingo-Enrich New York University cd2754@nyu.edu Yair Schiff Cornell University yzs2@cornell.edu Youssef Mroueh IBM Research AI mroueh@us.ibm.com
Pseudocode Yes Algorithm 1 given in App. F summarizes learning with the surrogate VDC. Algorithm 2 given in App. F summarizes learning with the surrogate CT distance.
Open Source Code Yes Code to reproduce experimental results is available here.
Open Datasets Yes In our experiments, we use following publicly available data: (1) the CIFAR-10 (Krizhevsky & Hinton, 2009) dataset, released under the MIT license, and (2) the Github icon silhouette, which was copied from https://github.com/CW-Huang/CP-Flow/blob/main/imgs/github.png.
Dataset Splits Yes We use the CIFAR-10 training data and split it as 95% training and 5% validation.
Hardware Specification Yes Training the baseline g0 and g with the surrogate VDC was done on a compute environment with 1 CPU and 1 A100 GPU.
Software Dependencies No Our experiments rely on various open-source libraries, including pytorch (Paszke et al., 2019) (license: BSD) and pytorch-lightning (Falcon et al., 2019) (Apache 2.0). The paper mentions software names but does not provide specific version numbers for them.
Experiment Setup Yes We set λ in Equation (11) to 10... We use ADAM optimizers (Kingma & Ba, 2015) for both networks, learning rates of 1e 4, and a batch size of 64.