Bayesian Inference for Structured Spike and Slab Priors

Authors: Michael R Andersen, Ole Winther, Lars K. Hansen

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using numerical experiments on synthetic data, we demonstrate the benefits of the model. [...] This section describes a series of numerical experiments that have been designed and conducted in order to investigate the properties of the proposed algorithm.
Researcher Affiliation Academia Michael Riis Andersen, Ole Winther & Lars Kai Hansen DTU Compute, Technical University of Denmark DK-2800 Kgs. Lyngby, Denmark {miri, olwi, lkh}@dtu.dk
Pseudocode Yes The proposed algorithm is summarized in figure 2. [Figure 2: Proposed algorithm for approximating the joint posterior distribution over x, z and γ.]
Open Source Code No The paper does not provide concrete access to source code for the methodology.
Open Datasets No The paper uses synthetic data and refers to existing concepts like 'EEG source localization with synthetic sources [22]' and 'Shepp-Logan Phantom experiment from [2]', but it does not provide any specific links, DOIs, repositories, or formal citations for publicly available datasets used in its experiments.
Dataset Splits No The paper describes how problem instances were generated and evaluated, but it does not specify explicit training, validation, or test dataset splits (e.g., percentages or counts).
Hardware Specification No The paper does not provide any specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library names with version numbers) needed to replicate the experiment.
Experiment Setup Yes We generate 100 problem instances from y = Ax0 + e, where the solutions vectors have been sampled from the proposed prior using the kernel Σi,j = 50 exp( ||i j||2 2/(2 102)), but constrained to have a fixed sparsity level of the K/D = 0.25. [...] The elements of A RN 250 are i.i.d Gaussian and the columns of A have been scaled to unit ℓ2-norm. The SNR is fixed at 20d B. [...] For the structured spike and slab method, we consider three different covariance structures: Σij = κ δ(i j), Σij = κ exp( ||i j||2/s) and Σij = κ exp( ||i j||2 2/(2s2)) with parameters κ = 50 and s = 10. In each case, we use a R = 50 rank approximation of Σ. [...] AEEG R128 800 is now a submatrix of a real EEG forward matrix corresponding to the grey area on the figure. The condition number of AEEG is 8 · 1015. The true sources X0 R800 20 are sampled from the structured spike and slab prior in eq. (8) using a squared exponential kernel with parameters A = 50, s = 10 and T = 20. The number of active sources is 46, i.e. x has 46 non-zero rows. SNR is fixed to 20d B.