Supervised Heterogeneous Domain Adaptation via Random Forests
Authors: Sanatan Sukhija, Narayanan C Krishnan, Gurkanwal Singh
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on three diverse datasets of varying dimensions and sparsity to verify the superiority of the proposed approach over other baseline and state of the art transfer approaches. |
| Researcher Affiliation | Academia | 1Department of Computer Science and Engineering, Indian Institute of Technology Ropar, Punjab, India sanatan@iitrpr.ac.in, ckn@iitrpr.ac.in 2Department of Computer Science and Engineering, PEC University of Technology, Chandigarh, India gurkanwal.singh7@gmail.com |
| Pseudocode | Yes | Algorithm 1 Supervised HDA via Random Forests (SHDA-RF) |
| Open Source Code | No | The paper does not provide concrete access to source code for the described methodology. No repository links or explicit statements about code availability are found. |
| Open Datasets | Yes | The CASAS dataset [Cook et al., 2013a] is a collection of smart home datasets that are widely used for investigating activity recognition algorithms. The 20 Newsgroups [Lang, 1995] text collection is a sparse dataset... The Statlog (Landsat Satellite) [Lichman, 2013] image dataset comprises of 6 classes and 36 real-valued features. |
| Dataset Splits | Yes | The target training set consists of approximately 7000 samples that preserve the original class distribution. 16 such random subsets are used for evaluating the performance of the different algorithms. Target training data is created by randomly selecting 10 samples per class. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details with version numbers, such as library or framework versions, needed to replicate the experiment. |
| Experiment Setup | Yes | The number of trees in the random forest was set to 100. The number of bagged features for learning in a tree in the forest was set to d + 5, where d is the total number of features. The parameters for the SVM model with RBF kernel were fine-tuned using grid search. Based on cross validation experiments, the length of ECOC was set to 35, beyond which the performance plateaued. |