@article{chen_dutton_ramachandra_wu_vatsavai_2021, title={Local Clustering with Mean Teacher for Semi-supervised learning}, ISSN={["1051-4651"]}, DOI={10.1109/ICPR48806.2021.9412469}, abstractNote={The Mean Teacher (MT) model of Tarvainen and Valpola has shown good performance on several semi-supervised benchmark datasets. MT maintains a teacher model's weights as the exponential moving average of a student model's weights and minimizes the divergence between their probability predictions under diverse perturbations of the inputs. However, MT is known to suffer from confirmation bias, that is, reinforcing incorrect teacher model predictions. In this work, we propose a simple yet effective method called Local Clustering (LC) to mitigate the effect of confirmation bias. In MT, each data point is considered independent of other points during training; however, data points are likely to be close to each other in feature space if they share similar features. Motivated by this, we cluster data points locally by minimizing the pairwise distance between neighboring data points in feature space. Combined with a standard classification cross-entropy objective on labeled data points, the misclassified unlabeled data points are pulled towards high-density regions of their correct class with the help of their neighbors, thus improving model performance. We demonstrate on semi-supervised benchmark datasets SVHN and CIFAR-10 that adding our LC loss to MT yields significant improvements compared to MT and performance comparable to the state of the art in semi-supervised learning11The code is available at: https://github.com/jay1204/local_clustering_with_mt_for_ssl.}, journal={2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)}, author={Chen, Zexi and Dutton, Benjamin and Ramachandra, Bharathkumar and Wu, Tianfu and Vatsavai, Ranga Raju}, year={2021}, pages={6243–6250} }
 @article{ramachandra_dutton_vatsavai_2019, title={Anomalous cluster detection in spatiotemporal meteorological fields}, volume={12}, ISSN={["1932-1872"]}, DOI={10.1002/sam.11398}, abstractNote={Finding anomalous regions in spatiotemporal climate data is an important problem with a need for greater accuracy. The collective and contextual nature of anomalies (e.g., heat waves) coupled with the real‐valued, seasonal, multimodal, highly correlated, and gridded nature of climate variable observations poses a multitude of challenges. Existing anomaly detection methods have limitations in the specific setting of real‐valued areal spatiotemporal data. In this paper, we develop a method for extreme event detection in meteorological datasets that follows from well known distribution‐based anomaly detection approaches. The method models spatial and temporal correlations explicitly through a piecewise parametric assumption and generalizes the Mahalanobis distance across distributions of different dimensionalities. The result is an effective method to mine contiguous spatiotemporal anomalous regions from meteorological fields which improves upon the current standard approach in climatology. The proposed method has been evaluated on a real global surface temperature dataset and validated using historical records of extreme events.}, number={2}, journal={STATISTICAL ANALYSIS AND DATA MINING}, author={Ramachandra, Bharathkumar and Dutton, Benjamin and Vatsavai, Ranga Raju}, year={2019}, month={Apr}, pages={88–100} }