Uncertainty quantification (UQ) plays a pivotal role in reduction of uncertainties during both optimization and decision making processes. It can be applied to solve a variety of real-world applications in science and engineering. Bayesian approximation and ensemble learning techniques are two most widely-used UQ methods in the literature. In this regard, researchers have proposed different UQ methods and examined their performance in a variety of applications such as computer vision (e.g., self-driving cars and object detection), image processing (e.g., image restoration), medical image analysis (e.g., medical image classification and segmentation), natural language processing (e.g., text classification, social media texts and recidivism risk-scoring), bioinformatics, etc. This study reviews recent advances in UQ methods used in deep learning. Moreover, we also investigate the application of these methods in reinforcement learning (RL). Then, we outline a few important applications of UQ methods. Finally, we briefly highlight the fundamental research challenges faced by UQ methods and discuss the future research directions in this field.

A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges

Buildings are among the largest consumers of electricity in the US. A significant portion of this energy use in buildings can be attributed to HVAC systems used to maintain comfort for occupants. In most cases these building HVAC systems run on fixed schedules and do not employ any fine grained control based on detailed occupancy information. In this paper we present the design and implementation of a presence sensor platform that can be used for accurate occupancy detection at the level of individual offices. Our presence sensor is low-cost, wireless, and incrementally deployable within existing buildings. Using a pilot deployment of our system across ten offices over a two week period we identify significant opportunities for energy savings due to periods of vacancy. Our energy measurements show that our presence node has an estimated battery lifetime of over five years, while detecting occupancy accurately. Furthermore, using a building simulation framework and the occupancy information from our testbed, we show potential energy savings from 10% to 15% using our system.

[ACM Press the 2nd ACM Workshop - Zurich, Switzerland (2010.11.02-2010.11.02)] Proceedings of the 2nd ACM Workshop on Embedded Sensing Systems for Energy-Efficiency in Building - BuildSys \'10 - Occupancy-driven energy management for smart building automation

Active learning (AL) attempts to maximize the performance gain of the model by marking the fewest samples. Deep learning (DL) is greedy for data and requires a large amount of data supply to optimize massive parameters, so that the model learns how to extract high-quality features. In recent years, due to the rapid development of internet technology, we are in an era of information torrents and we have massive amounts of data. In this way, DL has aroused strong interest of researchers and has been rapidly developed. Compared with DL, researchers have relatively low interest in AL. This is mainly because before the rise of DL, traditional machine learning requires relatively few labeled samples. Therefore, early AL is difficult to reflect the value it deserves. Although DL has made breakthroughs in various fields, most of this success is due to the publicity of the large number of existing annotation datasets. However, the acquisition of a large number of high-quality annotated datasets consumes a lot of manpower, which is not allowed in some fields that require high expertise, especially in the fields of speech recognition, information extraction, medical images, etc. Therefore, AL has gradually received due attention. A natural idea is whether AL can be used to reduce the cost of sample annotations, while retaining the powerful learning capabilities of DL. Therefore, deep active learning (DAL) has emerged. Although the related research has been quite abundant, it lacks a comprehensive survey of DAL. This article is to fill this gap, we provide a formal classification method for the existing work, and a comprehensive and systematic overview. In addition, we also analyzed and summarized the development of DAL from the perspective of application. Finally, we discussed the confusion and problems in DAL, and gave some possible development directions for DAL.

A Survey of Deep Active Learning

The classic supervised machine learning paradigm is based on learning in isolation, a single predictive model for a task using a single dataset. This approach requires a large number of training examples and performs best for well-defined and narrow tasks. Transfer learning refers to a set of methods that extend this approach by leveraging data from additional domains or tasks to train a model with better generalization properties. Over the last two years, the field of Natural Language Processing (NLP) has witnessed the emergence of several transfer learning methods and architectures which significantly improved upon the state-of-the-art on a wide range of NLP tasks. These improvements together with the wide availability and ease of integration of these methods are reminiscent of the factors that led to the success of pretrained word embeddings and ImageNet pretraining in computer vision, and indicate that these methods will likely become a common tool in the NLP landscape as well as an important research direction. We will present an overview of modern transfer learning methods in NLP, how models are pre-trained, what information the representations they learn capture, and review examples and case studies on how these models can be integrated and adapted in downstream NLP tasks.

Transfer Learning in Natural Language Processing.

Demand-side management technology is a key element of the proposed smart grid, which will help utilities make more efficient use of their generation assets by reducing consumers’ energy demand during peak load periods. However, although some modern appliances can respond to price signals from the utility companies, there is a vast stock of older appliances that cannot. For such appliances, utilities must infer what appliances are operating in a home, given only the power signals on the main feeder to the home (i.e., the home’s power consumption must be disaggregated into individual appliances). We report on an in-depth investigation of multi-label classification algorithms for disaggregating appliances in a power signal. A systematic review of this research topic shows that this class of algorithms has received little attention in the literature, even though it is arguably a more natural fit to the disaggregation problem than the traditional single-label classifiers used to date. We examine a multi-label meta-classification framework (RA ${k}$ EL), and a bespoke multi-label classification algorithm (ML ${k}$ NN), employing both time-domain and wavelet-domain feature sets. We test these classifiers on two real houses from the Reference Energy Disaggregation Dataset. We found that the multilabel algorithms are effective and competitive with published results on the datasets.

Toward Non-Intrusive Load Monitoring via Multi-Label Classification

We investigate the problem of making human activity recognition (AR) scalable-i.e., allowing AR classifiers trained in one context to be readily adapted to a different contextual domain. This is important because AR technologies can achieve high accuracy if the classifiers are trained for a specific individual or device, but show significant degradation when the same classifier is applied context-e.g., to a different device located at a different on-body position. To allow such adaptation without requiring the onerous step of collecting large volumes of labeled training data in the target domain, we proposed a transductive transfer learning model that is specifically tuned to the properties of convolutional neural networks (CNNs). Our model, called HDCNN, assumes that the relative distribution of weights in the different CNN layers will remain invariant, as long as the set of activities being monitored does not change. Evaluation on real-world data shows that HDCNN is able to achieve high accuracy even without any labeled training data in the target domain, and offers even higher accuracy (significantly outperforming competitive shallow and deep classifiers) when even a modest amount of labeled training data is available.

https://ink.library.smu.edu.sg/cgi/viewcontent.cgi?article=4979&context=sis_research

Scaling Human Activity Recognition via Deep Learning-based Domain Adaptation

Activity recognition in smart environment has been investigated rigorously in recent years. Researchers are enhancing the underlying activity discovery and recognition process by adding various dimensions and functionalities. But one significant barrier still persists which is collecting the ground truth information. Ground truth is very important to initialize a supervised learning of activities. Due to a large variety in number of Activities of Daily Living (ADLs), acknowledging them in a supervised way is a non-trivial research problem. Most of the previous researches have referenced a subset of ADLs and to initialize their model, they acquire a vast amount of informative labeled training data. On the other hand to collect ground truth and differentiate ADLs, human intervention is indispensable. As a result it takes an immense effort and raises privacy concerns to collect a reasonable amount of labeled data. In this paper, we propose to use active learning to alleviate the labeling effort and ground truth data collection in activity recognition pipeline. We investigate and analyze different active learning strategies to scale activity recognition and propose a dynamic k-means clustering based active learning approach. Experimental results on real data traces from a retirement community-(IRB #HP-00064387) help validate the early promise of our approach.

/pdf/active-learning-enabled-activity-recognition-24e3vt14qn.pdf

Active learning enabled activity recognition

Activity recognition using smartphone has great potential in many applications like healthcare, obesity management, abnormal behavior detection, public safety and security etc. Typical activity detection systems are built on to recognize a limited set of activities that are present in the training and testing environments. However, these systems require similar data distributions, activity sets and sufficient labeled training data in both training and testing phases. Therefore, inferring new activities is challenging in practical scenarios where training and testing environments are volatile, data distributions are diverge and testing environment has new set of activities with limited training samples. The shortage of labeled training data samples also degrades the activity recognition performance. In this work, we address these challenges by augmenting the Instance based Transfer Boost algorithm with k-means clustering. We evaluated our TransAct model with three public datasets - HAR, MHealth and DailyAndSports and demonstrated that our TransAct model outperforms traditional activity recognition approaches. Our experimental results show that our TransAct model achieves ≈ 81% activity detection accuracy on average.

TransAct: Transfer learning enabled activity recognition

Fine-grained monitoring of everyday appliances can provide better feedback to the consumers and motivate them to change behavior in order to reduce their energy usage. It also helps to detect abnormal power consumption events, long-term appliance malfunctions and potential safety concerns. Commercially available plug meters can be used for individual appliance monitoring but for an entire house, each such individual plug meters are expensive and tedious to setup. Alternative methods relying on Non-Intrusive Load Monitoring techniques help disaggregate electricity consumption data and learn about the individual appliance's power states and signatures. However fine-grained events (e.g., appliance malfunctions, abnormal power consumption, etc.) remain undetected and thus inferred contexts (such as safety hazards etc.) become invisible. In this work, we correlate an appliance's inherent acoustic noise with its energy consumption pattern individually and in presence of multiple appliances. We initially investigate classification techniques to establish the relationship between appliance power and acoustic states for efficient energy disaggregation and abnormal events detection. While promising, this approach fails when there are multiple appliances simultaneously in ‘ON’ state. To further improve the accuracy of our energy disaggregation algorithm, we propose a probabilistic graphical model, based on a variation of Factorial Hidden Markov Model (FHMM) for multiple appliances energy disaggregation. We combine our probabilistic model with the appliances acoustic analytics and postulate a hybrid model for energy disaggregation. Our approach helps to improve the performance of energy disaggregation algorithms and provide critical insights on appliance longevity, abnormal power consumption, consumer behavior and their everyday lifestyle activities. We evaluate the performance of our proposed algorithms on real data traces and show that the fusion of acoustic and power signatures can successfully detect a number of appliances with 95% accuracy.

Acoustic based appliance state identifications for fine-grained energy analytics

Activity recognition has applications in a variety of human-in-the-loop settings such as smart home health monitoring, green building energy and occupancy management, intelligent transportation, and participatory sensing. While fine-grained activity recognition systems and approaches help enable a multitude of novel applications, discovering them with non-intrusive ambient sensor systems pose challenging design, as well as data processing, mining, and activity recognition issues. In this paper, we develop a low-cost heterogeneous Radar based Activity Monitoring (RAM) system for recognizing fine-grained activities. We exploit the feasibility of using an array of heterogeneous micro-doppler radars to recognize low-level activities. We prototype a short-range and a long-range radar system and evaluate the feasibility of using the system for fine-grained activity recognition. In our evaluation, using real data traces, we show that our system can detect fine-grained user activities with 92.84% accuracy.

/pdf/ram-radar-based-activity-monitor-1eeohln761.pdf

Abdullah Al Hafiz Khan

Papers

Scaling Human Activity Recognition via Deep Learning-based Domain Adaptation

Active learning enabled activity recognition

TransAct: Transfer learning enabled activity recognition

Acoustic based appliance state identifications for fine-grained energy analytics

RAM: Radar-based activity monitor