The reliable and stable operation of modern shipboard power systems (SPSs) is crucial for mission success, particularly when faced with high-power pulsed loads and complex system interactions. While nonlinear model predictive control (NMPC) offers effective stabilization and robustness to bounded disturbances via its variant quasi-infinite-horizon NMPC (QIH-NMPC), it imposes a computational burden that limits real-time applicability for mission-critical SPSs at the microseconds level. This paper proposes a robust data-driven predictive control (DDPC) approach that leverages deep neural networks to approximate the QIH-NMPC control law, enabling fast-response and automated power management for medium-voltage direct current (MVDC) shipboard microgrids (MGs) that integrate synchronous generators, batteries, and supercapacitors. The proposed method includes a rigorous control design with theoretical guarantees and is validated under out-of-distribution (OOD) scenarios, aggressive load fluctuations with stochastic noise, stealthy false data injection attacks via numerical simulations, and fault conditions in a real-time simulator. Across diverse unseen disturbances, DDPC yields an average inference time of 6.91 μs, preserves control robustness by stabilizing the system within 0.05 seconds. It achieves a 97.8% reduction in computational load and improves control smoothness by 31% compared with NMPC, while also converging 50% faster than proportional-integral (PI) control. These findings highlight DDPC’s potential for automated, resilient, and real-time SPS control.

Genome sequencing technologies have the potential to transform clinical decision making and biomedical research by enabling high-throughput measurements of the genome at a granular level. However, to truly understand mechanisms of disease and predict the effects of medical interventions, high-throughput data must be integrated with demographic, phenotypic, environmental, and behavioral data from individuals. Further, effective knowledge discovery methods must infer relationships between these data types. We recently proposed a pipeline (CausalMGM) to achieve this. CausalMGM uses probabilistic graphical models to infer the relationships between variables in the data; however, CausalMGM's graphical structure learning algorithm can only handle small datasets efficiently. We propose a new methodology (piPref-Div) that selects the most informative variables for CausalMGM, enabling it to scale. We validate the efficacy of piPref-Div against other feature selection methods and demonstrate how the use of the full pipeline improves breast cancer outcome prediction and provides biologically interpretable views of gene expression data.

This paper presents a structure generation algorithm, which converts rough human drawings into stable structures comprising rectangular blocks, suitable for physics-based 2-D environments. Generating viable structures for a physics-based environment imposes many additional requirements above those of most traditional sketch-based domains. Our method is sophisticated enough to deal with these requirements, while still ensuring that the generated structure accurately represents the original sketch. We describe and implement a framework for this process, allowing inexperienced users to create complex structures with ease. Multiple structure possibilities are identified for a single drawing and are then compared based on their similarity to the original sketch using a heuristic value. We evaluate our approach by investigating its ability to replicate structures for the video game Angry Birds, based on human drawn sketches of the original levels.

Driven by the increasing gap between the exponential growth of data and the limited human ability to comprehend them, recently, a novel interactive data exploration approach called Explore-by-Examples has generated a lot of attention for its capabilities to bridge this gap and to help the user obtain high-value content from the data that are often hidden using the traditional search methods. However, despite their effectiveness in extracting valuable information, existing Explore-by-Examples systems focus solely on structured data, which represents a small portion of the data available today. In this work, we present a novel data exploration framework, namely ExNav (Exploration Navigator), which enables the user to effortlessly explore the world of unstructured data for insights that are often unreachable from traditional search and exploration methods. In particular, we exploit the space of advanced machine learning, data embedding, and active learning algorithms to design effective exploration and space pruning approaches tailored for unstructured datasets. Our experimental evaluation using multiple real-world unstructured datasets (i.e., text, image, and graph) show that ExNav can reduce users' effort by up to 9x while still achieving the same accuracy as the state-of-the-art alternative. Moreover, ExNav is also able to identify relevant data items that are often undetectable by current techniques, even when a large number of samples are explored.

In this paper we present several proofs for the computational complexity of the physics-based video game Angry Birds. We are able to demonstrate that solving levels for different versions of Angry Birds is either NP-hard, PSPACE-hard, PSPACE-complete or EXPTIME-hard, depending on the maximum number of birds available and whether the game engine is deterministic or stochastic. We believe that this is the first time that a single-player video game has been proven EXPTIME-hard.

Existing visualization recommendation approaches introduced an overwhelmingly large number of utility functions (UFs) to rank the visualizations (i.e., views). In order to discover the ideal UF and their tunable parameters that are most suited for a particular analysis context, recent works have proposed Interactive View Recommendation (IVR). IVR identifies the ideal UF by learning from the interactions with the user during the analysis process. We claim that the user is usually unable to provide an accurate real number between 0 and 1, indicating the view interestingness, and in this work, propose three alternative feedback types: binary, Likert scale, and pairwise comparison. We further developed ranking algorithms for the three feedback types. Finally, we propose different example selection strategies, and experimentally evaluate them on the three feedback types. We found that uncertainty-based and hybrid-based strategies usually outperforms the random strategy in terms of recommendation accuracy across all feedback types, while the interestingness-based strategy usually performs worse than the random strategy.

The CAPRIOv2.0 system is the evolution of CAPRIO, our context-aware path recommendation system that has as its primary objectives the minimum outdoor exposure and distance of the recommended path. CAPRIOv2.0 offers enhanced indoor context-awareness in terms of accessibility and congestion. In this demonstration, we exhibit CAPRIOv2.0 and present its novel graph representation that integrates accessibility, congestion, indoor, and outdoor information to discover paths satisfying accessibility, outdoor exposure, and distance constraints of an individual. We further present a new spatial model index, called SMI-tree, which enables CAPRIO v2.0 to quickly forecast the congestion in corridors and hallways. Individuals can interactively engage with the CAPRIOv2.0 GUI using any of their devices to appreciate how our proposed structures and algorithms can provide an alternative context-aware path by combining outdoor, indoor, congestion and accessibility information.Video https://db.cs.pitt.edu/caprio/v2.

Traditional venue and tour recommendation systems do not necessarily provide a diverse set of recommendations and leave little room for serendipity . In this article, we design MPG, a Mobile Personal Guide that recommends: (i) a set of diverse yet surprisingly interesting venues that are aligned to user preferences and (ii) a set of routes, constructed from the recommended venues. We also introduce EPUI, an Experimental Platform for Urban Informatics. Our comparison with the state-of-the-art schemes indicates that MPG is capable of providing high-quality venues and route recommendations while incorporating seamlessly both the notion of diversity and that of serendipity.

During extreme weather conditions and natural disasters caused by meteorological phenomena, it is imperative to enable navigation that minimizes the outdoor section of recommended paths. Existing indoor-outdoor navigation and localization systems have evolved to support queries like the shortest distance, either outdoor or indoor, with additional constraints. However, most of them work in isolation and do not take into consideration the external natural conditions, like the weather, that an individual may experience walking outside during a polar vortex or heatwave. In this paper, we present CAPRIO, a context-aware path recommendation system whose objectives are two-fold: (i) minimizing outdoor exposure; and (ii) minimizing the distance of the recommended path. We propose a novel graph representation that integrates indoor and outdoor information to discover paths that satisfy outdoor exposure and distance constraints. We measure the efficiency of the proposed solution using two real datasets collected from the University of Pittsburgh and University of Cyprus campuses. We show that we can achieve comparable distance to the state-of-the-art in minimizing outdoor exposure.

Recently, navigation and localization systems have emerged to support queries like the shortest distance in either indoor or outdoor with additional constraints. These systems, however, neither combine the indoor and outdoor information nor consider the external natural conditions like the weather that one may face across an outdoor path. In this demonstration paper we present CAPRIO , which proposes and implements a novel graph representation that integrates indoor and outdoor information to discover paths that personalize outdoor exposure while minimizes the overall path length. We also demonstrate how unifying the graph algorithms for indoor and outdoor navigation enables significant optimizations that would not be possible otherwise.

This paper presents an overview of the second AIBIRDS level generation competition, held jointly at the 2017 IEEE Conference on Computational Intelligence and Games and the 26th International Joint Conference on Artificial Intelligence. This competition tasked entrants with developing a level generator for the physics-based puzzle game Angry Birds. Submitted generators were required to deal with many physical reasoning constraints caused by the realistic nature of the game's environment, in addition to ensuring that the created levels were fun, challenging, and solvable. This year's competition was a significant improvement over the previous year, with a greater number of participants and more advanced generators. In this paper, we describe the framework, rules, submitted generators, and results for this competition. We also provide some background information on related research and other video game AI competitions and discuss what can be learned from this year's competition. There are several game and real-world applications for this type of research, and we provide some examples of the types of levels we would like future competition entries to generate.

Recent studies in urban navigation have revealed new demands (e.g., diversity, safety, happiness, serendipity) for the navigation services that are critical to providing useful recommendations to travelers. This exposes the need to design next-generation navigation services that accommodate these newly emerging aspects. In this paper, we present a prototype system, namely, EPUI (an Experimental Platform of Urban Informatics), which provides a testbed for exploring and evaluating venues and route recommendation solutions that balance between different objectives (i.e., demands) including the newly discovered ones. In addition, EPUI incorporates a modularized design, enabling researchers to upload their own algorithms and compare them to well-known algorithms using different performance metrics. Its user interface makes it easily usable by both end-user and experienced researchers.

In this demo we present a prototype of an experimental platform for evaluating item recommendation algorithms. The application domain for our system is that of digital city guides. Our prototype implementation allows the user to explore different algorithms and compare their output. Among the algorithms implemented is MPG, which aims at providing a diverse set of recommendations better aligned with user preferences. MPG takes into consideration the user preferences (e.g., reach willing to cover, types of venues interested in exploring etc.), the popularity of the establishments as well as their distance from the current location of the user by combining them into a single composite score. We provide a web interface, which outputs on a map the recommended locations along with metadata (e.g., type and name of location, relevance and diversity scores, etc.). It also illustrates the potential of the Preferential Diversity approach on which MPG is based.

With the proliferation of mobile computing and the ability to collect detailed data for the urban environment a number of systems that aim at providing Points of Interest (POIs) and tour recommendations have appeared. The overwhelming majority of these systems aims at providing an optimal recommendation, where optimality refers to objectives of minimizing the distance to be covered or maximizing the quality of the POIs recommended. A major problem is that by focusing on the optimization of these objectives, there remains little room to the user for serendipity. Urban and social scientists have identified serendipity, i.e., the ability to come across unexpected places, as a feature that makes a city livable. In this work, we introduce a prototype of an experimental platform for evaluating venue recommendation algorithms by providing informative tour recommendations based on the suggested venues. Our prototype system integrates the notion of serendipity in urban navigation at both the venue as well as the route recommendation level without compromising the quality and diversity of the recommended POIs. In addition, our system allows the user to upload their own algorithms and explore their performance as compared to many well-known algorithms.

The exponential growth of high dimensional biological data has led to a rapid increase in demand for automated approaches for knowledge production. Existing methods rely on two general approaches to address this challenge: 1) the Theory-driven approach, which utilizes prior accumulated knowledge, and 2) the Data-driven approach, which solely utilizes the data to deduce scientific knowledge. Both of these approaches alone suffer from bias toward past/present knowledge, as they fail to incorporate all of the current knowledge that is available to make new discoveries. In this paper, we show how an integrated method can effectively address the high dimensionality of big biological data, which is a major problem for pure data-driven analysis approaches. We realize our approach in a novel two-step analytical workflow that incorporates a new feature selection paradigm as the first step to handling high-throughput gene expression data analysis and that utilizes graphical causal modeling as the second step to handle the automatic extraction of causal relationships. Our results, on real-world clinical datasets from The Cancer Genome Atlas (TCGA), demonstrate that our method is capable of intelligently selecting genes for learning effective causal networks.

The Angry Birds AI Competition (aibirds.org) has been held annually since 2012 in conjunction with some of the major AI conferences, most recently with IJCAI 2015. The goal of the competition is to build AI agents that can play new Angry Birds levels as good as or better than the best human players. Successful agents should be able to quickly analyze new levels and to predict physical consequences of possible actions in order to select actions that solve a given level with a high score. Agents have no access to the game internal physics, but only receive screenshots of the live game. In this paper we describe why this problem is a challenge for AI, and why it is an important step towards building AI that can successfully interact with the real world. We also summarise some highlights of past competitions, including a new competition track we introduced recently.

Exploration over large datasets is a key first step in data analysis, as users may be unfamiliar with the underlying database schema and unable to construct precise queries that represent their interests. Such data exploration task usually involves executing numerous ad-hoc queries, which requires a considerable amount of time and human effort. In this paper, we present REQUEST, a novel framework that is designed to minimize the human effort and enable both effective and efficient data exploration. REQUEST supports the query-from-examples style of data exploration by integrating two key components: 1) Data Reduction, and 2) Query Selection. As instances of the REQUEST framework, we propose several highly scalable schemes, which employ active learning techniques and provide different levels of efficiency and effectiveness as guided by the user's preferences. Our results, on real-world datasets from Sloan Digital Sky Survey, show that our schemes on average require 1-2 orders of magnitude fewer feedback questions than the random baseline, and 3-16× fewer questions than the state-of-the-art, while maintaining interactive response time. Moreover, our schemes are able to construct, with high accuracy, queries that are often undetectable by current techniques.

The capability of determining the right sequence of physical actions to achieve a given task is essential for AI that interacts with the physical world. The great difficulty in developing this capability has two main causes: (1) the world is continuous and therefore the action space is infinite, (2) due to noisy perception, we do not know the exact physical properties of our environment and therefore cannot precisely simulate the consequences of a physical action.

The proliferation of mobile, ubiquitous and spatial computing has led to a number of services aiming into facilitate the exploration of a city. Platforms such as Foursquare and Yelp curate information about establishments in an area that can then be used for recommendation purposes. Traditionally an approach followed by these systems is to rank places based on their popularity, proximity or any other feature that represents the quality of the venue and then return the top-k of them. However, this approach, while simple and intuitive, is not necessarily providing a diverse set of recommendations, since similar venues typically are ranked closely. Therefore, in this paper we design and introduce MPG (which stands for Mobile Personal Guide), a mobile service that provides a set of diverse venue recommendations better aligned with user preferences. MPG takes into consideration the user preferences (e.g., distance willing to cover, types of venues interested in exploring, etc.), the popularity of the establishments, as well as their distance from the current location of the user by combining them in a single composite score. We evaluate our approach using a large-scale dataset of approximately 14 million venues collected from Foursquare. Our results indicate that MPG can increase coverage of the result set compared to the baselines considered. It further achieves a significantly better Relevancy-Diversity trade-off ratio.

Many current computer vision approaches for object detection can only detect objects that have been learned in advance. In this paper, we present a method that uses qualitative stability analysis to infer the existence of unknown objects in certain areas of the images based on gravity and stability of already detected objects. Our method recursively searches these areas for unknown objects until all detected objects form a stable structure or no new objects can be identified anymore. We evaluate our method using the popular video game Angry Birds. We only start with detecting the green pigs and are able to automatically identify and detect all essential game objects in all 400+ available levels. All objects can be accurately and reliably detected. Our method can be applied to other video games where objects obey gravity and are bound by polygons.