Transfer Learning for Latent Variable Network Models

Authors: Akhil Jalan, Arya Mazumdar, Soumendu Sundar Mukherjee, Purnamrita Sarkar

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we empirically demonstrate our algorithm s use on real-world and simulated network estimation problems.
Researcher Affiliation Academia Akhil Jalan Department of Computer Science UT Austin akhiljalan@utexas.edu Arya Mazumdar Halıcıo glu Data Science Institute & Dept of CSE UC San Diego arya@ucsd.edu Soumendu Sundar Mukherjee Statistics and Mathematics Unit (SMU) Indian Statistical Institute, Kolkata ssmukherjee@isical.ac.in Purnamrita Sarkar Department of Statistics and Data Sciences UT Austin purna.sarkar@austin.utexas.edu
Pseudocode Yes Algorithm 1 b Q-Estimation for Latent Variable Models
Open Source Code Yes We submit our code as a supplementary zip file in accordance with the Neur IPS code and data submission guidelines.
Open Datasets Yes Metabolic Networks. We access metabolic models from King et al. (2016) at http://bigg.ucsd.edu. (...) EMAIL-EU. We use the email-EU-core-temporal dataset at https://snap.stanford.edu/data/email-Eu-core-temporal.html, as introduced in Paranjape et al. (2017).
Dataset Splits No The paper describes using source and target data for estimation, but does not explicitly mention or specify training, validation, and test dataset splits or cross-validation procedures for model evaluation.
Hardware Specification No As described in Appendix C, 'We run all experiments on a personal Linux machine with 378GB of CPU/RAM.' This description does not include specific CPU or GPU models, or other detailed hardware specifications.
Software Dependencies No The paper does not explicitly list specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions, or specific libraries).
Experiment Setup Yes Hyperparameters. We do not tune any hyperparameters. For Algorithm 1 we use the quantile cutoff hn = qn Q in all experiments.