Learning Neurosymbolic Generative Models via Program Synthesis

Authors: Halley Young, Osbert Bastani, Mayur Naik

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we evaluate our approach on synthetic data and on a real-world dataset of building facades (Tyleˇcek & ˇS ara, 2013), both on the task of generation from scratch and on generation from a partial image. We show that our approach substantially outperforms several state-of-the-art deep generative models (Section 4).
Researcher Affiliation Academia 1University of Pennsylvania, USA. Correspondence to: Halley Young <halleyy@seas.upenn.edu>.
Pseudocode Yes Algorithm 1 Synthesizes a program P representing the global structure of a given image x 2 RNM NM.
Open Source Code No The paper does not provide any specific repository link or explicit statement about the release of the source code for the methodology described.
Open Datasets Yes Synthetic dataset. We developed a synthetic dataset based on MNIST.
Dataset Splits No The paper states dataset sizes for training and testing (e.g., '10,000 training and 500 test images' and '1755 training, 100 testing') but does not specify a separate validation split.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper discusses various model architectures and frameworks (e.g., LSTM, VAE, Cycle GAN, GLCIC) but does not provide specific software names with version numbers (e.g., PyTorch 1.9, TensorFlow 2.x, or specific Python library versions).
Experiment Setup No The paper mentions neural network architectures and general training strategies but lacks specific experimental setup details such as concrete hyperparameter values (e.g., learning rate, batch size, number of epochs) or optimizer settings in the main text or appendices.