Latent Normalizing Flows for Many-to-Many Cross-Domain Mappings

Authors: Shweta Mahajan, Iryna Gurevych, Stefan Roth

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of our model on diverse tasks, including image captioning and text-to-image synthesis. To that end, we perform experiments on the COCO dataset (Lin et al., 2014).
Researcher Affiliation Academia Shweta Mahajan, Iryna Gurevych, Stefan Roth Department of Computer Science, TU Darmstadt, Germany
Pseudocode No The paper provides architectural diagrams and mathematical formulations but does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements about releasing source code or providing links to a code repository.
Open Datasets Yes To that end, we perform experiments on the COCO dataset (Lin et al., 2014).
Dataset Splits Yes It contains 82,783 training and 40,504 validation images, each with five captions. Following Wang et al. (2016); Mao et al. (2015) for image captioning, we use 118,287 data points for training and evaluate on 1,000 test images.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU models, CPU types, memory) used to run the experiments.
Software Dependencies No The paper mentions software components like VGG-16, GRU, and LSTM in the architecture details but does not provide specific version numbers for any libraries, frameworks, or programming languages used.
Experiment Setup Yes The overall objective of our semi-supervised generative model framework to be minimized is given by Lµ(xt,xv) = λ1DKL(qθ1 zs|xt, xv) pφs(zs) + λ2DKL(qθ2 z t|xt, zs) pφt(z t|zs) + λ3DKL qθ3(z v|xv, zs) pφv(z v|zs) + λ4Lrec t (xt, xt) + λ5Lrec v (xv, xv).