Sharp Bounds for Generalized Causal Sensitivity Analysis

Authors: Dennis Frauen, Valentyn Melnychuk, Stefan Feuerriegel

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

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we propose a scalable algorithm to estimate our sharp bounds from observational data and perform extensive computational experiments to show the validity of our bounds empirically.
Researcher Affiliation Academia Dennis Frauen, Valentyn Melnychuk & Stefan Feuerriegel LMU Munich Munich Center for Machine Learning {frauen,melnychuk,feuerriegel}@lmu.de
Pseudocode Yes Algorithm 1: Causal sensitivity analysis with mediators
Open Source Code Yes 1Code is available at https://github.com/DennisFrauen/SharpCausalSensitivity.
Open Datasets Yes The preprocessed data is publically available at https://github.com/jopersson/covid19-mobility/blob/main/Data.
Dataset Splits No We perform hyperparameter tuning for our experiments on synthetic data using grid search on a validation set. ... The paper states that a validation set was used for hyperparameter tuning but does not provide specific details on the split percentage or methodology.
Hardware Specification No The paper does not specify any hardware components (e.g., CPU, GPU models, memory) used for the experiments.
Software Dependencies No We use feed-forward neural networks with softmax activation function... For densities, we use conditional normalizing flows [73] (neural spline flows [21])... We perform training using the Adam optimizer [41]. ... The paper mentions various software components and libraries (e.g., neural networks, normalizing flows, Adam optimizer) but does not provide specific version numbers for them.
Experiment Setup Yes We perform hyperparameter tuning for our experiments on synthetic data using grid search on a validation set. The tunable parameters and search ranges are shown in Table 2. ... Table 2: MODEL TUNABLE PARAMETERS SEARCH RANGE includes details for Epochs, Batch size, Learning rate, Hidden layer size, Number of spline bins, and Dropout probability.