More Like this

1. UnfoldML: Cost-Aware and Uncertainty-Based Dynamic 2D Prediction for Multi-Stage Classification

   Xu, Yanbo; Khare, Alind; Matlin, Glenn; Ramadoss, Monish; Kamaleswaran, Rishikesan; Zhang, Chao; Tumanov, Alexey (January 2022, Advances in neural information processing systems)

   Koyejo, S.; Mohamed, S.; Agarwal, A.; Belgrave, D.; Cho, K.; Oh, A. (Ed.)

   Machine Learning (ML) research has focused on maximizing the accuracy of predictive tasks. ML models, however, are increasingly more complex, resource intensive, and costlier to deploy in resource-constrained environments. These issues are exacerbated for prediction tasks with sequential classification on progressively transitioned stages with “happens-before” relation between them.We argue that it is possible to “unfold” a monolithic single multi-class classifier, typically trained for all stages using all data, into a series of single-stage classifiers. Each single- stage classifier can be cascaded gradually from cheaper to more expensive binary classifiers that are trained using only the necessary data modalities or features required for that stage. UnfoldML is a cost-aware and uncertainty-based dynamic 2D prediction pipeline for multi-stage classification that enables (1) navigation of the accuracy/cost tradeoff space, (2) reducing the spatio-temporal cost of inference by orders of magnitude, and (3) early prediction on proceeding stages. UnfoldML achieves orders of magnitude better cost in clinical settings, while detecting multi- stage disease development in real time. It achieves within 0.1% accuracy from the highest-performing multi-class baseline, while saving close to 20X on spatio- temporal cost of inference and earlier (3.5hrs) disease onset prediction. We also show that UnfoldML generalizes to image classification, where it can predict different level of labels (from coarse to fine) given different level of abstractions of a image, saving close to 5X cost with as little as 0.4% accuracy reduction. 

   more » « less
2. Image classification and training with severe data loss

   https://doi.org/10.1117/12.2633172  

   Marquard, Dillon; Wright, Kyle; Marcia, Roummel F. (October 2022, SPIE Optical Engineering + Applications, 2022)

   Zelinski, Michael E.; Taha, Tarek M.; Howe, Jonathan (Ed.)

   Image classification forms an important class of problems in machine learning and is widely used in many realworld applications, such as medicine, ecology, astronomy, and defense. Convolutional neural networks (CNNs) are machine learning techniques designed for inputs with grid structures, e.g., images, whose features are spatially correlated. As such, CNNs have been demonstrated to be highly effective approaches for many image classification problems and have consistently outperformed other approaches in many image classification and object detection competitions. A particular challenge involved in using machine learning for classifying images is measurement data loss in the form of missing pixels, which occurs in settings where scene occlusions are present or where the photodetectors in the imaging system are partially damaged. In such cases, the performance of CNN models tends to deteriorate or becomes unreliable even when the perturbations to the input image are small. In this work, we investigate techniques for improving the performance of CNN models for image classification with missing data. In particular, we explore training on a variety of data alterations that mimic data loss for producing more robust classifiers. By optimizing the categorical cross-entropy loss function, we demonstrate through numerical experiments on the MNIST dataset that training with these synthetic alterations can enhance the classification accuracy of our CNN models. 

   more » « less
3. Learning-Based Workload Phase Classification and Prediction Using Performance Monitoring Counters

   https://doi.org/10.1109/MLCAD52597.2021.9531161  

   Alcorta, Erika S.; Gerstlauer, Andreas (August 2021, ACM/IEEE Workshop on Machine Learning for CAD (MLCAD))

   null (Ed.)

   Predicting coarse-grain variations in workload behavior during execution is essential for dynamic resource optimization of processor systems. Researchers have proposed various methods to first classify workloads into phases and then learn their long-term phase behavior to predict and anticipate phase changes. Early studies on phase prediction proposed table-based phase predictors. More recently, simple learning-based techniques such as decision trees have been explored. However, more recent advances in machine learning have not been applied to phase prediction so far. Furthermore, existing phase predictors have been studied only in connection with specific phase classifiers even though there is a wide range of classification methods. Early work in phase classification proposed various clustering methods that required access to source code. Some later studies used performance monitoring counters, but they only evaluated classifiers for specific contexts such as thermal modeling. In this work, we perform a comprehensive study of source-oblivious phase classification and prediction methods using hardware counters. We adapt classification techniques that were used with different inputs in the past and compare them to state-of-the-art hardware counter based classifiers. We further evaluate the accuracy of various phase predictors when coupled with different phase classifiers and evaluate a range of advanced machine learning techniques, including SVMs and LSTMs for workload phase prediction. We apply classification and prediction approaches to SPEC workloads running on an Intel Core-i9 platform. Results show that a two-level kmeans clustering combined with SVM-based phase change prediction provides the best tradeoff between accuracy and long-term stability. Additionally, the SVM predictor reduces the average prediction error by 80% when compared to a table-based predictor. 

   more » « less
4. Real-Time Posture Classification on Resource-Constrained Runtime Environments: A Comparative Study of Traditional and Deep Learning Models

   https://doi.org/10.1109/SoutheastCon56624.2025.10971620  

   Shejuti, Zarin; Fuad, Muztaba (March 2025, IEEE)

   While machine learning models perform well on offline data, assessing their performance in real-world, resource-constrained environments-considering accuracy, prediction time, power consumption, and memory usage-is crucial for practical applications. This research implements a mobile-based Human Activity Recognition solution to classify three postures-sitting, standing, and walking-using smartphone sensors, specifically accelerometer, gyroscope, and magnetometer. Time-domain features extracted from these sensors were used, with Random Forest employed for feature selection. One traditional machine learning model, Logistic Regression, and one deep learning model, Convolutional Neural Network, were trained and deployed via an Android application for real-time evaluation. While the Convolutional Neural Network achieved higher accuracy and better memory efficiency, Logistic Regression demonstrated faster prediction times during real-time use. Both models showed reduced accuracy for standing and walking postures in real-world conditions, emphasizing the challenges of deploying machine learning models in dynamic environments. This study highlights the importance of evaluating machine learning models in real-world settings to ensure reliability and efficiency, particularly in resource-constrained environments. 

   more » « less
5. Adversarial Attack on Graph Structured Data

   Dai, Hanjun; Li, Hui; Tian, Tian; Huang, Xin; Wang, Lin; Zhu, Jun; Song, Le (July 2018, Proceedings of the 35th International Conference on Machine Learning, PMLR)

   Deep learning on graph structures has shown exciting results in various applications. However, few attentions have been paid to the robustness of such models, in contrast to numerous research work for image or text adversarial attack and defense. In this paper, we focus on the adversarial attacks that fool deep learning models by modifying the combinatorial structure of data. We first propose a reinforcement learning based attack method that learns the generalizable attack policy, while only requiring prediction labels from the target classifier. We further propose attack methods based on genetic algorithms and gradient descent in the scenario where additional prediction confidence or gradients are available. We use both synthetic and real-world data to show that, a family of Graph Neural Network models are vulnerable to these attacks, in both graph-level and node-level classification tasks. We also show such attacks can be used to diagnose the learned classifiers. 

   more » « less