Active Nearest Neighbors in Changing Environments
Authors: Christopher Berlind, Ruth Urner
ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments show that ANDA successfully corrects for dataset bias in multiclass image categorization. Our experiments on synthetic data illustrate ANDA s adaptation ability and show that its classification performance compares favorably with baseline passive nearest neighbors. Experiments on challenging image classification tasks show that ANDA is a good candidate for correcting dataset bias. |
| Researcher Affiliation | Academia | Christopher Berlind CBERLIND@GATECH.EDU Georgia Institute of Technology, Atlanta, GA, USA; Ruth Urner RUTH.URNER@TUEBINGEN.MPG.DE Max Planck Institute for Intelligent Systems, T ubingen, Germany |
| Pseudocode | Yes | Algorithm 1 ANDA: Active NN Domain Adaptation; Algorithm 2 Safe: Find a (k, k )-NN-cover; Algorithm 3 EMMA: Efficient multiset multicover approximation for finding a small (k, k )-NN-cover |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | Tommasi & Tuytelaars (2014) aligned and preprocessed several image datasets that provide a way of comparing domain adaptation methods on this problem. Of the four datasets (Caltech256, Imagenet, Bing, and SUN) we chose not to use SUN because the differences in how data was labeled result in a clear violation of covariate shift. |
| Dataset Splits | No | The paper mentions 'unlabeled target sample sizes' and 'test sets' but does not specify a separate validation split or explicit percentages/counts for training, validation, and testing. It discusses averaging over trials and using test sets but lacks explicit validation split information. |
| Hardware Specification | No | The paper does not provide any specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers. |
| Experiment Setup | Yes | For all algorithms and target sample sizes we fixed m S = 3200, k = 7, and k = 21. (Synthetic Data); For all algorithms and target sample sizes we fixed m S = 2000, k = 25, and k = 75. (Image Classification) |