Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Sensing Theorems for Unsupervised Learning in Linear Inverse Problems

Authors: Julián Tachella, Dongdong Chen, Mike Davies

JMLR 2023 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform a series of numerical experiments to illustrate the theoretical bounds presented in Sections 3 and 4. ... Figure 8 shows the probability of recovery over 25 Monte Carlo trials for different numbers of measurements m and operators |G|. ... Figure 11a shows the average test peak-signal-to-noise ratio (PSNR) achieved by the trained model for |G| = 1, 10, 20, 30, 40 and m = 1, 100, 200, 300, 400. ... We use the standard MNIST dataset...
Researcher Affiliation Academia Juli an Tachella EMAIL Laboratoire de Physique CNRS, ENSL Lyon, F-69364, France; Dongdong Chen EMAIL School of Engineering University of Edinburgh Edinburgh, EH9 3FB, UK; Mike Davies EMAIL School of Engineering University of Edinburgh Edinburgh, EH9 3FB, UK
Pseudocode No The paper describes algorithms in section 6 'Algorithms' using paragraph text, but does not present any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain an explicit statement about the release of source code for the methodology described, nor does it provide a link to a code repository.
Open Datasets Yes We use the standard MNIST dataset which has an approximate box-counting dimension k = 12 (Hein and Audibert, 2005). The dataset contains N = 60000 training samples...
Dataset Splits Yes The dataset contains N = 60000 training samples, and these are partitioned such that N/|G| different samples are observed via each operator. The test set consists of 10000 samples, which are also randomly divided into |G| parts, one per operator.
Hardware Specification No The paper does not provide specific details regarding the hardware used to run its experiments, such as GPU or CPU models.
Software Dependencies No The paper does not provide specific software dependencies or version numbers for the tools and libraries used in the implementation.
Experiment Setup Yes The networks are trained using the Adam optimizer. ... we use an autoencoder architecture with 3 hidden layers with 1000, 32 and 1000 neurons, as shown in Figure 10. We use relu non-linearities between layers, except at the output of the last layer.