Variational Autoencoders for Highly Multivariate Spatial Point Processes Intensities

Authors: Baichuan Yuan, Xiaowei Wang, Jianxin Ma, Chang Zhou, Andrea L. Bertozzi, Hongxia Yang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show the method s utility on both synthetic data and real-world data sets.
Researcher Affiliation Collaboration Baichuan Yuan1, Xiaowei Wang2, Jianxin Ma2, Chang Zhou2, Andrea L. Bertozzi1, Hongxia Yang2 1Department of Mathematics, University of California, Los Angeles 2DAMO Academy, Alibaba Group
Pseudocode Yes Algorithm 1: Training VAE SPP with stochastic gradient descent.
Open Source Code No The paper states 'We implement our models in Tensorflow based on VAE-CF2', with footnote 2 pointing to 'https://github.com/dawenl/vae_cf'. This link is to a third-party VAE-CF implementation, not the authors' own VAE-SPP code or an explicit statement of its release.
Open Datasets Yes We consider the Gowalla data set (Cho et al., 2011) in New York City (NYC) and California (CA). ... Movie Lens data sets (ML-100K and ML-1M) include the movie (item) rating by users
Dataset Splits Yes We split the data into training, validation and testing sets. ... We randomly select 500 users as the validation set and 500 users as the testing set. ... We set the size of both validation and testing sets to 100.
Hardware Specification Yes We conducted the experiments on a single GTX 1080 TI 11GB GPU.
Software Dependencies No The paper mentions 'Tensorflow', 'python statsmodel', and 'GPy' but does not provide specific version numbers for any of these software dependencies.
Experiment Setup Yes For simulation data, we train both models for 200 epochs using Adam optimizer with β = 0.2, lr = 5 10 5. We use mini-batches of size 20. Our architectures consist of a one layer MLP with K = 50. For VAE-SPP, σ2 = 0.001.