Kevin Miller’s research while affiliated with Brigham Young University and other places

We develop a contrastive graph-based active learning pipeline (CGAP) to identify surface water and nearwater sediment pixels in multispectral images. CGAP enhances the graph-based active learning pipeline (GAP) designed for surface water and sediment detection in multispectral imagery (10.1109/IGARSS52108.2023.10282009), which outperformed conventional methods such as CNN-Unet, SVM, and RF. Our improvements focus on boosting both the pipeline’s robustness and efficiency by integrating a feature-embedding neural network prior to graph construction. Trained using contrastive learning, this neural network projects high-dimensional raw features into a lower-dimensional space, facilitating more efficient graph learning. The training process incorporates specialized augmentations to bolster the embedded features’ resilience to geometric transformations, varying resolutions, and light cloud cover. Moreover, we develop a Python-based demo, GraphRiverClassifier (GRC), that uses Google Earth Engine and our enhanced pipeline to provide a user-friendly tool for rapid and accurate surface water and sediment analyses and rapid testing of algorithm performances.

Active learning in semi-supervised classification involves introducing additional labels for unlabelled data to improve the accuracy of the underlying classifier. A challenge is to identify which points to label to best improve performance while limiting the number of new labels. “Model Change” active learning quantifies the resulting change incurred in the classifier by introducing the additional label(s). We pair this idea with graph-based semi-supervised learning (SSL) methods, that use the spectrum of the graph Laplacian matrix, which can be truncated to avoid prohibitively large computational and storage costs. We consider a family of convex loss functions for which the acquisition function can be efficiently approximated using the Laplace approximation of the posterior distribution. We show a variety of multiclass examples that illustrate improved performance over prior state-of-art.

We develop a contrastive graph-based active learning pipeline (CGAP) to identify surface water and near-water sediment pixels in multispectral images. CGAP enhances the graph-based active learning pipeline (GAP) (10.1109/IGARSS52108.2023.10282009), which outperforms methods such as CNN-Unet, support vector machine (SVM), and random forest (RF), while requiring less training data. Active learning plays an important role for training data reduction, resulting in an order of magintude less training data compared with conventional methods and three or more orders of magnitude less compared with CNN-Unet. Our improvements focus on boosting both the pipeline's robustness and efficiency by integrating a feature-embedding neural network prior to graph construction. This neural network, trained using contrastive learning, performs effective data dimension reduction by projecting high-dimensional raw features into a lower-dimensional space, thereby facilitating more efficient graph learning. The training process incorporates specialized augmentations to bolster the embedded features' resilience to geometric transformations, varying resolutions, and light cloud cover. Moreover, we develop a Python-based demo, GraphRiverClassifier (GRC), that uses the Google Earth Engine and our enhanced pipeline to provide a user-friendly tool for rapid and accurate surface water and sediment analyses and rapid testing of algorithm performances.

Graph learning, when used as a semi-supervised learning (SSL) method, performs well for classification tasks with a low label rate. We provide a graph-based batch active learning pipeline for pixel/patch neighborhood multi- or hyperspectral image segmentation. Our batch active learning approach selects a collection of unlabeled pixels that satisfy a graph local maximum constraint for the active learning acquisition function that determines the relative importance of each pixel to the classification. This work builds on recent advances in the design of novel active learning acquisition functions (e.g., the Model Change approach in arXiv:2110.07739) while adding important further developments including patch-neighborhood image analysis and batch active learning methods to further increase the accuracy and greatly increase the computational efficiency of these methods. In addition to improvements in the accuracy, our approach can greatly reduce the number of labeled pixels needed to achieve the same level of the accuracy based on randomly selected labeled pixels.

Despite the empirical success and practical significance of (relational) knowledge distillation that matches (the relations of) features between teacher and student models, the corresponding theoretical interpretations remain limited for various knowledge distillation paradigms. In this work, we take an initial step toward a theoretical understanding of relational knowledge distillation (RKD), with a focus on semi-supervised classification problems. We start by casting RKD as spectral clustering on a population-induced graph unveiled by a teacher model. Via a notion of clustering error that quantifies the discrepancy between the predicted and ground truth clusterings, we illustrate that RKD over the population provably leads to low clustering error. Moreover, we provide a sample complexity bound for RKD with limited unlabeled samples. For semi-supervised learning, we further demonstrate the label efficiency of RKD through a general framework of cluster-aware semi-supervised learning that assumes low clustering errors. Finally, by unifying data augmentation consistency regularization into this cluster-aware framework, we show that despite the common effect of learning accurate clusterings, RKD facilitates a "global" perspective through spectral clustering, whereas consistency regularization focuses on a "local" perspective via expansion.

Active learning improves the performance of machine learning methods by judiciously selecting a limited number of unlabeled data points to query for labels, with the aim of maximally improving the underlying classifier's performance. Recent gains have been made using sequential active learning for synthetic aperture radar (SAR) data arXiv:2204.00005. In each iteration, sequential active learning selects a query set of size one while batch active learning selects a query set of multiple datapoints. While batch active learning methods exhibit greater efficiency, the challenge lies in maintaining model accuracy relative to sequential active learning methods. We developed a novel, two-part approach for batch active learning: Dijkstra's Annulus Core-Set (DAC) for core-set generation and LocalMax for batch sampling. The batch active learning process that combines DAC and LocalMax achieves nearly identical accuracy as sequential active learning but is more efficient, proportional to the batch size. As an application, a pipeline is built based on transfer learning feature embedding, graph learning, DAC, and LocalMax to classify the FUSAR-Ship and OpenSARShip datasets. Our pipeline outperforms the state-of-the-art CNN-based methods.

We develop a graph active learning pipeline (GAP) to detect surface water and in-river sediment pixels in satellite images. The active learning approach is applied within the training process to optimally select specific pixels to generate a hand-labeled training set. Our method obtains higher accuracy with far fewer training pixels than both standard and deep learning models. According to our experiments, our GAP trained on a set of 3270 pixels reaches a better accuracy than the neural network method trained on 2.1 million pixels.