S$Ω$I: Score-based O-INFORMATION Estimation

Authors: Mustapha Bounoua, Giulio Franzese, Pietro Michiardi

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments validate our approach on synthetic data, and demonstrate the effectiveness of SΩI in the context of a real-world use case.
Researcher Affiliation Collaboration 1Ampere Software Technology, France 2Department of Data Science, Eurecom, France.
Pseudocode Yes Algorithm 1 SΩI Training step
Open Source Code Yes We provide code-base for SΩI implementation at 1. The training of SΩI is carried out using Adam optimizer (Kingma & Ba, 2015). 1https://github.com/Mustapha Bounoua/soi
Open Datasets Yes We consider the Visual Behavior project, which used the Allen Brain Observatory to collect a highly standardized dataset consisting of recordings of neural activity in mice (Allen-Institute, 2022).
Dataset Splits No For each experiment, we use 100k samples for training the various neural estimators, and 10k samples at inference time, to estimate O-INFORMATION. The paper explicitly mentions training and testing sets, but a separate validation split is not specified.
Hardware Specification No No specific hardware details such as GPU or CPU models, memory, or cluster specifications are provided in the paper.
Software Dependencies No The paper mentions software components like 'Adam optimizer' and frameworks like 'VP-SDE formulation', but it does not specify exact version numbers for any of them (e.g., PyTorch 1.9, Python 3.8).
Experiment Setup Yes For our method SΩI, we use the VP-SDE formulation (Song et al., 2021) and learn a unique denoising network to estimate the various score terms. The denoiser is a simple, stacked multilayer perceptron (MLP) with skip connections, adapted to the input dimension. We apply importance sampling (Huang et al., 2021; Song et al., 2021) at both training and inference time. Finally, we use 10-sample Monte Carlo estimates for computing integrals. (Table 1 provides hyperparameters like Width, Time embed, Batch size, Lr, Iterations).