Discriminative Metric Learning by Neighborhood Gerrymandering

Authors: Shubhendu Trivedi, David Mcallester, Greg Shakhnarovich

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In experiments on a variety of data sets our method is shown to achieve excellent results compared to current state of the art in metric learning.
Researcher Affiliation Academia Shubhendu Trivedi, David Mc Allester, Gregory Shakhnarovich Toyota Technological Institute Chicago, IL 60637 {shubhendu,mcallester,greg}@ttic.edu
Pseudocode Yes Algorithm 1: Stochastic gradient descent; Algorithm 2: Targeted inference; Algorithm 3: Loss augmented inference
Open Source Code No The paper does not provide an explicit statement or a link to open-source code for the described methodology.
Open Datasets Yes We compare the error of k NN classifiers using metrics learned with our approach to that with other learned metrics. For this evaluation we replicate the protocol in [11], using the seven data sets in Table 1.
Dataset Splits Yes for the other five data sets, we report the mean and standard errors of 5-fold cross validation (results for all methods are on the same folds). [...] The value of C is tuned on on a 75%/25% split of the training portion.
Hardware Specification No The paper mentions running experiments but does not specify any hardware components (e.g., CPU, GPU models, memory).
Software Dependencies No The paper does not mention any specific software dependencies with version numbers.
Experiment Setup Yes Our SGD algorithm stops when the running average of the surrogate loss over most recent epoch no longer descreases substantially, or after max. number of iterations. We use learning rate η(t) = 1/t. [...] The value of C is tuned on on a 75%/25% split of the training portion.