Emerging AI systems will be making more and more decisions that impact the lives of humans in a significant way. It is essential, then, that these AI systems make decisions that take into account the desires, goals, and preferences of other people, while simultaneously learning about what those preferences are. In this work, we argue that the reinforcement-learning framework achieves the appropriate generality required to theorize about an idealized ethical artificial agent, and offers the proper foundations for grounding specific questions about ethical learning and decision making that can promote further scientific investigation. We define an idealized formalism for an ethical learner, and conduct experiments on two toy ethical dilemmas, demonstrating the soundness and flexibility of our approach. Lastly, we identify several critical challenges for future advancement in the area that can leverage our proposed framework.

The Natural Language Processing, Artificial Intelligence, and Robotics fields have made significant progress towards developing robust component technologies (speech recognition/synthesis, machine translation, image recognition); advanced inference mechanisms that accommodate uncertainty and noise; and autonomous driving systems that operate seamlessly on our roads. In spite of this, we do not yet know how to talk to the machines we build or have them speak to us in natural language; how to make them smarter via simple, natural language instructions; how to understand what they are about to do; or how to work with them collaboratively towards accomplishing some joint goal. In this paper, we discuss our work towards building a dataset that enables an empirical approach to studying the relation between natural language, actions, and plans; and introduce a problem formulation that allows us to take meaningful steps towards addressing the open problems listed above.

Wireless sensor networks play a major role in smart grids and smart buildings. They are not just used for sensing, but they are also used as actuating. In terms of sensing they are used to measure temperature, humidity, light, to detect motion, etc. Sensors are often operated on a battery and hence we often face a trade-off between obtaining frequent sensor readings versus maximizing their battery life. There have been several approaches to maximizing their battery life from hardware level to software level such as reducing components energy consumption, limiting node operation capabilities, using power-aware routing protocols, and adding solar energy support. In this paper, we introduce a novel approach: we model the sensor readings in a wireless network using a dynamic Gaussian Bayesian network (dGBn) whose structure is automatically learned from data. dGBn allows us to integrate information across sensors and infer missing readings more accurately. Through active inference for dGBns, we are able to actively choose which sensors should be pulled for a reading and which ones can stay in a power-saving mode at each time step, maximizing prediction accuracy while staying within the budgetary constraints on battery consumption.

In intensive care units (ICU), electrocardiogram (ECG) waveforms show diverse variationsunder different patients' physical conditions.In general, physicians can diagnose patients efficientlyby detecting any disorder of heart rate or rhythm and any change in the morphological pattern of ECG data,which contain underlying semantics.To help physicians better analyze ECG data in a fairly short time,it is essential to develop a novel method for mining semantics from ECG patterns.This paper is the very first time to characterize ECG patterns by using Prefix Scalable Pattern Tree (PSP-Tree).Comparing with similar currently existing methods, PSP-Tree can mine significant semantics,such as scalability, temporality and hierarchy over ECG patterns.We conduct extensive experiments on real ECG data set which are obtained from PhysioBank Community and Beijing No.3 People Hospital.The experiment results show that our method performs more feasibly and effectively than other related work.

In this paper we approach a subset of semantic analysis tasks through a symbiotic cognitive computing approach -- the user and the system learn from each other and accomplish the tasks better than they would do on their own. Our approach starts with a domain expert building a simplified domain model (e.g. semantic lexicons) and annotating documents with that model. The system helps the user by allowing them to obtain quicker results, and by leading them to refine their understanding of the domain. Meanwhile, through the feedback from the user, the system adapts more quickly and produces more accurate results. We believe this virtuous cycle is key for building next generation high quality semantic analysis systems. We present some preliminary findings and discuss our results on four aspects of this virtuous cycle, namely: the intrinsic incompleteness of semantic models, the need for a human in the loop, the benefits of a computer in the loop and finally the overall improvements offered by the human-computer interaction in the process.

Utilities use Demand Response (DR) to balance supply and demand in the electric grid by involving customers in efforts to reduce electricity consumption during peak periods. To implement and adapt DR under dynamically changing conditions of the grid, reliable prediction of reduced consumption is critical. However, despite the wealth of research on electricity consumption prediction and DR being long in practice, the problem of reduced consumption prediction remains largely un-addressed. In this paper, we identify unique computational challenges associated with the prediction of reduced consumption and contrast this to that of normal consumption and DR baseline prediction. We propose a novel ensemble model that leverages different sequences of daily electricity consumption on DR event days as well as contextual attributes for reduced consumption prediction. We demonstrate the success of our model on a large, real-world, high resolution dataset from a university microgrid comprising of over 950 DR events across a diverse set of 32 buildings. Our model achieves an average error of 13.5%, an 8.8% improvement over the baseline. Our work is particularly relevant for buildings where electricity consumption is not tied to strict schedules. Our results and insights should prove useful to the researchers and practitioners working in the sustainable energy domain.

In this paper we explain our concept of Interaction and Task Patterns, and discuss how such patterns can be applied to support mixed-initiative in symbiotic human-robot interaction both with service and industrial robotic systems.

We propose a new model for learning bilingual word representations from non-parallel document-aligned data. Following the recent advances in word representation learning, our model learns dense real-valued word vectors, that is, bilingual word embeddings (BWEs). Unlike prior work on inducing BWEs which heavily relied on parallel sentence-aligned corpora and/or readily available translation resources such as dictionaries, the article reveals that BWEs may be learned solely on the basis of document-aligned comparable data without any additional lexical resources nor syntactic information. We present a comparison of our approach with previous state-of-the-art models for learning bilingual word representations from comparable data that rely on the framework of multilingual probabilistic topic modeling (MuPTM), as well as with distributional local context-counting models. We demonstrate the utility of the induced BWEs in two semantic tasks: (1) bilingual lexicon extraction, (2) suggesting word translations in context for polysemous words. Our simple yet effective BWE-based models significantly outperform the MuPTM-based and context-counting representation models from comparable data as well as prior BWE-based models, and acquire the best reported results on both tasks for all three tested language pairs.

Model-Based Diagnosis (MBD) is a principled and domain-independent way of analyzing why a system under examination is not behaving as expected. Given an abstract description (model) of the system's components and their behavior when functioning normally, MBD techniques rely on observations about the actual system behavior to reason about possible causes when there are discrepancies between the expected and observed behavior. Due to its generality, MBD has been successfully applied in a variety of application domains over the last decades. In many application domains of MBD, testing different hypotheses about the reasons for a failure can be computationally costly, e.g., because complex simulations of the system behavior have to be performed. In this work, we therefore propose different schemes of parallelizing the diagnostic reasoning process in order to better exploit the capabilities of modern multi-core computers. We propose and systematically evaluate parallelization schemes for Reiter's hitting set algorithm for finding all or a few leading minimal diagnoses using two different conflict detection techniques. Furthermore, we perform initial experiments for a basic depth-first search strategy to assess the potential of parallelization when searching for one single diagnosis. Finally, we test the effects of parallelizing "direct encodings" of the diagnosis problem in a constraint solver.

The paper focuses on finding the m best solutions to combinatorial optimization problems using best-first or depth-first branch and bound search. Specifically, we present a new algorithm m-A*, extending the well-known A* to the m-best task, and for the first time prove that all its desirable properties, including soundness, completeness and optimal efficiency, are maintained. Since best-first algorithms require extensive memory, we also extend the memory-efficient depth-first branch and bound to the m-best task. We adapt both algorithms to optimization tasks over graphical models (e.g., Weighted CSP and MPE in Bayesian networks), provide complexity analysis and an empirical evaluation. Our experiments confirm theory that the best-first approach is largely superior when memory is available, but depth-first branch and bound is more robust. We also show that our algorithms are competitive with related schemes recently developed for the m-best task.