Learning with Operator-valued Kernels in Reproducing Kernel Krein Spaces

Authors: Akash Saha, Balamurugan Palaniappan

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments with indefinite operator-valued kernels on synthetic and real data sets demonstrate the utility of the proposed approach.
Researcher Affiliation Academia Akash Saha IEOR, IIT Bombay Mumbai, India akashsaha@iitb.ac.in P. Balamurugan IEOR, IIT Bombay Mumbai, India balamurugan.palaniappan@iitb.ac.in
Pseudocode Yes Algorithm 1: Op MINRES
Open Source Code Yes All methods were coded in Python 3.6 and the codes are made public1. 1The codes used for experiments can be found at https://github.com/akashsaha06/NeurIPS-2020/
Open Datasets Yes We use the Haskins IEEE Rate Comparison DB dataset available at https://yale.app.box.com/s/cfn8hj2puveo65fq54rp1ml2mk7moj3h (Tiede et al., 2017).
Dataset Splits Yes The experiments performed used 320 samples for training and 80 samples for testing. For hyperparameter tuning, we used 3-fold multi-grid cross validation for all the methods.
Hardware Specification No All experiments were run on a Linux box with 182 Gigabytes main memory and 28 CPU cores. While memory and core count are given, specific CPU models or types are not mentioned, which is required for detailed hardware specification.
Software Dependencies Yes All methods were coded in Python 3.6
Experiment Setup Yes λ was chosen from {10 3, 10 2, 0.1, 1, 10, 100}. γ, γ1, γ2, η, η1, η2 were chosen from {0.001, 0.002, . . . , 0.009, 0.01, 0.02, . . . , 0.09, 0.1, 0.2, . . . , 0.9, 1, 2, . . . , 10, 20, . . . , 100}.