Maximal Sparsity with Deep Networks?

Authors: Bo Xin, Yizhou Wang, Wen Gao, David Wipf, Baoyuan Wang

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply the proposed system to a practical photometric stereo computer vision problem, where the goal is to estimate the 3D geometry of an object using only 2D photos taken from a single camera under different lighting conditions. In this context, shadows and specularities represent sparse outliers that must be simultaneously removed from 104 106 surface points. We achieve state-of-the-art performance using only weak supervision despite a minuscule computational budget appropriate for real-time mobile environments. Synthetic Tests with Correlated Dictionaries: We generate a dictionary matrix Φ Rn m using Φ = Pn i=1 1 i2 uiv i , where ui Rn and vi Rm have iid elements drawn from N(0, 1). We also rescale each column of Φ to have unit ℓ2 norm.
Researcher Affiliation Collaboration Bo Xin1,2 Yizhou Wang1 Wen Gao1 Baoyuan Wang3 David Wipf2 1Peking University 2Microsoft Research, Beijing 3Microsoft Research, Redmond
Pseudocode No The paper describes algorithms using mathematical equations and descriptions but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statements about releasing code or links to source code repositories for the described methodology.
Open Datasets No Synthetic Tests with Correlated Dictionaries: We generate a dictionary matrix Φ Rn m using Φ = Pn i=1 1 i2 uiv i , where ui Rn and vi Rm have iid elements drawn from N(0, 1). ... As real-world training data is expensive to acquire, we instead use weak supervision by synthetically generating a training set as follows.
Dataset Splits No We used N1 = 600000 samples for training and the remaining N2 = 100000 for testing. A specific validation split is not explicitly mentioned.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running experiments.
Software Dependencies No The paper does not list specific software components with version numbers that would be needed for replication.
Experiment Setup No Detailed network design, optimization setup, and alternative metrics can be found in [26].