Learning with Group Invariant Features: A Kernel Perspective.

Authors: Youssef Mroueh, Stephen Voinea, Tomaso A. Poggio

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the validity of these claims on three datasets: text (artificial), vision (MNIST), and speech (TIDIGITS).
Researcher Affiliation Collaboration Youssef Mroueh IBM Watson Group mroueh@us.ibm.com Stephen Voinea CBMM, MIT. voinea@mit.edu Tomaso Poggio CBMM, MIT . tp@ai.mit.edu
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or link for the open-sourcing of the code for their described methodology.
Open Datasets Yes MNIST (Figure 2): We seek local invariance to translation and rotation, and so all random templates are translated by up to 3 pixels in all directions and rotated between -20 and 20 degrees. TIDIGITS (Figure 3): We use a subset of TIDIGITS consisting of 326 speakers (men, women, children) reading the digits 0-9 in isolation, and so each datapoint is a waveform of a single word.
Dataset Splits No The paper mentions 'training set' and 'test points' but does not specify training, validation, and test splits with percentages, absolute counts, or cross-validation details.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments.
Software Dependencies No All RLS experiments in this paper were completed with the GURLS toolbox [23]. (No version number provided for GURLS toolbox).
Experiment Setup Yes RLS will perform the optimization, min W Rm T 1 N ||Y Φ(X)W||2 F + λ||W||2 F , where || ||F is the Frobenius norm, λ is the regularization parameter, and Φ is the feature map, which for the representation described in this paper will be a CDF pooling of the data projected onto group-transformed random templates. [...] Φ = CDF(n, m) refers to a random feature map with n bins and m templates.