Convex Co-Embedding for Matrix Completion with Predictive Side Information

Authors: Yuhong Guo

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conducted experiments on two types of applications, transductive multi-label learning with incomplete labels and recommendation matrix completion. In this section we present the experimental settings and results.
Researcher Affiliation Academia Yuhong Guo School of Computer Science Carleton University, Ottawa, Canada yuhong.guo@carleton.ca
Pseudocode Yes Algorithm 1 Fast Proximal Gradient Descent Algorithm
Open Source Code No The paper does not provide any statement regarding the release of source code, nor does it include a link to a code repository.
Open Datasets Yes We conducted experiments for transductive incomplete multi-label learning on ten standard multi-label datasets for web page classification from yahoo.com (Ueda and Saito 2002). We have also conducted recommendation experiments on three real-word Amazon datasets: Beauty, Office and Sports.
Dataset Splits Yes For each dataset, we randomly sampled 10% instances for testing and used the remaining 90% data for training. We did parameter selection for the regularization parameter γ from the set 2{ 9, 8, ,8,9} by using two-fold cross-validation on the labeled training data.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific version numbers for any software dependencies or libraries used in the experiments.
Experiment Setup Yes The proposed approach, Co Embed, has three trade-off parameters α, γ and ρ. Note as shown in Proposition 1, any difference between α and γ will simply lead to rescaling the input feature values. Hence we just set α = γ. Moreover, the ρ parameter controls the degree of the soft approximation for the equality constraints. It should be a reasonably large value. In our experiments, we set ρ = 100. We did parameter selection for the regularization parameter γ from the set 2{ 9, 8, ,8,9} by using two-fold cross-validation on the labeled training data.