We present an electricity demand forecasting algorithm based on Gaussian processes. By introducing a task-specific, custom covariance function k_power, which incorporates all available seasonal information as well as weather data, we are able to make accurate predictions of power consumption and renewable energy production. The hyper-parameters of the Gaussian process are optimized automatically using marginal likelihood maximization. There are no parameters to be specified by the user. We evaluate the prediction performance on simulated data and get superior results compared to a simple baseline method.

The workshop on Trading Agent Design and Analysis focuses on all aspects of the design and evaluation of trading agents, including agent architectures, decision-making algorithms, theoretic analysis of agents or market games, empirical studies of agent performance, agent negotiation strategies, game-theoretic studies, market architectures and other related topics.

Mobile robots moving in a crowd need to conform to the same social standards as the human participants. Imitating human behavior is a natural choice in these situations — however, not every human behaves in the same way. On the other hand, it is known that humans tend to behave in a consistent way, with their behavior predictable by their social status. In this paper we consider a marketplace where humans perform purposeful movement. With many people moving on intersecting trajectories, the participants occasionally encounter em micro-conflicts, where they need to balance their desire to move towards their destination (their mission) with the requirements of the social norms of not bumping into strangers or violating their personal space. We model micro-conflicts by a series of two-player games. We show that if all humans are using consistent strategies which are aware of their own social status and can infer the social status of their opponent, the overall social costs will be lower compared to scenarios where the humans perform inconsistent strategies (even if those strategies are adaptive). We argue that robots acting in social environments should also adopt consistent strategies and align themselves with the ongoing social structure.

The advent and ubiquity of mass-market portable computational devices has opened up new opportunities for the development of assistive technologies for disabilities, especially within the domain of augmentative and alternative communications devices. Word prediction can facilitate everyday communication on mobile devices by reducing the physical interactions required to produce dialogue with them. To support personalized word prediction, a text prediction system should learn from the user’s own data to update the initial learned likelihoods that provide high quality "out of the box" performance. Within this lies an inherent trade-off: a larger corpus of initial training data can yield better default performance, but may also increase the amount of user data required for personalization of the system to be effective. We investigate a learning approach employing hierarchical modeling of phrases expected to offer sufficient "out of the box" performance relative to other learning approaches, while reducing the amount of initial training data required to facilitate on-line personalization of the text prediction system. The key insight of the proposed approach is the separation of stopwords, which primarily play syntactical roles in phrases, from keywords, which provide context and meaning in the phrase. This allows the abstraction of a phrase from an ordered list of all words to an ordered list of keywords. Thus the proposed hierarchical modeling of phrases employs two layers: keywords and stopwords. A third level abstracting keywords to a single topic is also considered, combining the power of both topic modeling and trigrams to make predictions within and between layers.

We propose using deep learning as the "workhorse" of a cognitive architecture. We show how deep learning can be leveraged to learn representations, such as a hierarchy of analogical schemas, from relational data. This approach to higher cognition drives some desiderata of deep learning, particularly modality independence and the ability to make top-down predictions. Finally, we consider the problem of how relational representations might be learned from sensor data that is not explicitly relational.

This position paper describes two approaches towards the representations that a robot can learn from its experience. In the first approach, the robot learns models for reasoning about human-interpretable aspects of the environment, for example models of space and objects. In the second approach, the robot incrementally learns predictions for the consequences of performing policies, where a policy is any experimental procedure that the robot can perform. These two approaches correspond closely to the ideas of a scientific model and an experimental prediction, and ideally the benefits of both can be accessible to a robot.

Getting an embodied cognitive system to form a mental model of its world is a challenging prospect. Most AI systems leverage domains defined entirely by the system designers—initial objects, relations, operations, and even search control knowledge are often pre-specified. Building autonomous systems that can bootstrap themselves using a minimal domain definition is a critical research objective of Developmental AI. PAM-P2, a domain agnostic cognitive system, builds a world model using an initial set of user specified primitive actions and homeostatic needs. As sensory datasets are received, an ontology is formed and used to derive situations, events, episodes, solutions, problems, and predictions. An overview of the PAM-P2 architecture and knowledge representation is presented.

The digitalization of music and the instant availability of millions of tracks on the Internet require new approaches to support the user in the exploration of these huge music collections. One possible approach to address this problem, which can also be found on popular online music platforms, is the use of user-created or automatically generated playlists (mixes). The automated generation of such playlists represents a particular type of the music recommendation problem with two special characteristics. First, the tracks of the list are usually consumed immediately at recommendation time; secondly, songs are listened to mostly in consecutive order so that the sequence of the recommended tracks can be relevant. In the past years, a number of different approaches for playlist generation have been proposed in the literature. In this paper, we review the existing core approaches to playlist generation, discuss aspects of appropriate offline evaluation designs and report the results of a comparative evaluation based on different datasets. Based on the insights from these experiments, we propose a comparably simple and computationally tractable new baseline algorithm for future comparisons, which is based on track popularity and artist information and is competitive with more sophisticated techniques in our evaluation settings.

Robust integration of robotic and human perception abilities can greatly enhance the execution of complex information-driven tasks like search and rescue. Our goal is to formally characterize and combine diverse information streams obtained from multiple autonomous robots and humans within a unified probabilistic framework that naturally supports autonomous perception, human situational awareness, and cooperative human-robot task execution under stochastic uncertainties. This approach requires well-designed human-robot interfaces, flexible and accurate probabilistic models for exploiting “human sensor” data, and sophisticated Bayesian inference methods for efficient learning and online dynamic state estimation. We review some of recent theoretical developments and insights from experiments using real human-robot teams, and discuss some open challenges for future research.

We have been successfully deploying mobile service robots in an office building to execute user-requested tasks, such as delivering messages, transporting items, escorting people, and enabling telepresence. Users submit task requests to a dedicated website which forms a schedule of tasks for the robots to execute. The robots autonomously navigate in the building to complete their tasks. However, upon observing the many successful task executions, we realized that the robots are too autonomous in their determination to execute a planned task, with no mechanism to interrupt or redirect the robot through local interaction. In this work, we analyze the challenges of this goal of interruption, and contribute an approach to interrupt the robot anywhere during its execution through spoken dialog. Tasks can then be modified or new tasks can be added through speech, allowing users to manage the robot’s schedule. We discuss the response of the robot to human interruptions. We also introduce a finite state machine based on spoken dialog to handle the conversations that might occur during task execution. The goal is for the robot to fulfill humans’ requests as much as possible while minimizing the impact to the ongoing and pending tasks. We present examples of our task interruption scheme executing on robots to demonstrate its effectiveness.