Structured Prediction with Projection Oracles

Authors: Mathieu Blondel

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate our losses on label ranking, ordinal regression and multilabel classification, confirming the improved accuracy enabled by projections.
Researcher Affiliation Industry Mathieu Blondel NTT Communication Science Laboratories Kyoto, Japan
Pseudocode No The paper describes methods textually but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes A Python implementation is available at https://github.com/mblondel/projection-losses.
Open Datasets Yes We use the same six public datasets as in [26].
Dataset Splits Yes choosing λ against the validation set.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for running experiments were found in the paper.
Software Dependencies No A Python implementation is available at https://github.com/mblondel/projection-losses.
Experiment Setup No In all cases, we use a linear model θ = g(x) := Wx and solve 1 n Pn i=1 SΨ C (Wxi, yi) + λ 2 W 2 F by L-BFGS, choosing λ against the validation set.