We present a system that produces sentential descriptions of video: who did what to whom, and where and how they did it. Action class is rendered as a verb, participant objects as noun phrases, properties of those objects as adjectival modifiers in those noun phrases, spatial relations between those participants as prepositional phrases, and characteristics of the event as prepositional-phrase adjuncts and adverbial modifiers. Extracting the information needed to render these linguistic entities requires an approach to event recognition that recovers object tracks, the track-to-role assignments, and changing body posture.

We present a framework that allows an observer to determine occluded portions of a structure by finding the maximum-likelihood estimate of those occluded portions consistent with visible image evidence and a consistency model. Doing this requires determining which portions of the structure are occluded in the first place. Since each process relies on the other, we determine a solution to both problems in tandem. We extend our framework to determine confidence of one's assessment of which portions of an observed structure are occluded, and the estimate of that occluded structure, by determining the sensitivity of one's assessment to potential new observations. We further extend our framework to determine a robotic action whose execution would allow a new observation that would maximally increase one's confidence.

The common internal structure and algorithmic organization of object detection, detection-based tracking, and event recognition facilitates a general approach to integrating these three components. This supports multidirectional information flow between these components allowing object detection to influence tracking and event recognition and event recognition to influence tracking and object detection. The performance of the combination can exceed the performance of the components in isolation. This can be done with linear asymptotic complexity.

In information retrieval, a fundamental goal is to transform a document into concepts that are representative of its content. The term "representative" is in itself challenging to define, and various tasks require different granularities of concepts. In this paper, we aim to model concepts that are sparse over the vocabulary, and that flexibly adapt their content based on other relevant semantic information such as textual structure or associated image features. We explore a Bayesian nonparametric model based on nested beta processes that allows for inferring an unknown number of strictly sparse concepts. The resulting model provides an inherently different representation of concepts than a standard LDA (or HDP) based topic model, and allows for direct incorporation of semantic features. We demonstrate the utility of this representation on multilingual blog data and the Congressional Record.

A few months ago, a graduate student in another country called me (Vaughan) to ask for the source code of one of my multi-robot simulation experiments. The student had an idea for a modification that she thought would improve the system's performance. By the standards of scientific practice this was a perfectly reasonable request and I felt obliged to give it to her. With our original code, the student could (i) rerun our experiments to verify that we reported the results correctly; (ii) inspect the code to make sure that it actually implements the algorithm described in our paper; (iii) change parameters and initial conditions to make sure our results were not a fluke of the particular experimental setting; (iv) modify the robot controllers and quantitatively compare her new method with our originals. It would cost me nothing to make her a copy of our code, and her methodology would be impeccable. Why then do we read so few papers using this methodology? It turned out to be impossible to identify exactly which code was used to perform the experiments in our years-old paper. We had not labeled the source code at that moment, and it had subsequently been modified. All the code was under version control, so we could obtain approximately the right code by looking at revision dates.

Relational data representations have become an increasingly important topic due to the recent proliferation of network datasets (e.g., social, biological, information networks) and a corresponding increase in the application of statistical relational learning (SRL) algorithms to these domains. In this article, we examine a range of representation issues for graph-based relational data. Since the choice of relational data representation--for the nodes, links, and features--can dramatically affect the capabilities of SRL algorithms, we survey approaches and opportunities for relational representation transformation designed to improve the performance of these algorithms. This leads us to introduce an intuitive taxonomy for data representation transformations in relational domains that incorporates link transformation and node transformation as symmetric representation tasks. In particular, the transformation tasks for both nodes and links include (i) predicting their existence, (ii) predicting their label or type, (iii) estimating their weight or importance, and (iv) systematically constructing their relevant features. We motivate our taxonomy through detailed examples and use it to survey and compare competing approaches for each of these tasks. We also discuss general conditions for transforming links, nodes, and features. Finally, we highlight challenges that remain to be addressed.

Targeting at sparse learning, we construct Banach spaces B of functions on an input space X with the properties that (1) B possesses an l1 norm in the sense that it is isometrically isomorphic to the Banach space of integrable functions on X with respect to the counting measure; (2) point evaluations are continuous linear functionals on B and are representable through a bilinear form with a kernel function; (3) regularized learning schemes on B satisfy the linear representer theorem. Examples of kernel functions admissible for the construction of such spaces are given.

The first indication of a new outbreak is often in unstructured data (natural language) and reported openly in traditional or social media as a new `flu-like' or `malaria-like' illness weeks or months before the new pathogen is eventually isolated. We present a system for tracking these early signals globally, using natural language processing and crowdsourcing. By comparison, search-log-based approaches, while innovative and inexpensive, are often a trailing signal that follow open reports in plain language. Concentrating on discovering outbreak-related reports in big open data, we show how crowdsourced workers can create near-real-time training data for adaptive active-learning models, addressing the lack of broad coverage training data for tracking epidemics. This is well-suited to an outbreak information-flow context, where sudden bursts of information about new diseases/locations need to be manually processed quickly at short notice.

Computational social science, as with all complex adaptive systems sciences, involves a great amount of uncertainty on several fronts, including intrinsic arbitrariness such as that due to path dependence, disagreement on social theory and how to capture it in software, input data of different credibility that does not exactly match the requirements of software because it was gathered for another purpose, and inexactly matching translations between models that were designed for different purposes than the study at hand. This paper presents a method of formally tracking that uncertainty, keeping the data input parameters proportionate with logical and probabilistic constraints, and capturing proportionate dynamics of the output ordered by the decision points of policy change, for the purpose of risk-based analysis. Once ordered this way, the data can be compared to other data similarly expressed, whether that data is from simulation excursions or from the real world, for objective comparison and distance scoring at the level of dynamic patterns as opposed to single outcome validation. This method enables wargame adjudicators to be run out with data gleaned from the wargame, enables data to be repurposed for both training and testing set, and facilitates objective validation scoring through soft matching. Artificial intelligence tools used in the method include probabilistic ontologies with crisp and Bayesian inference, game trees that are multiplayer non-zero sum and decision point based rather than turn-based, and Markov processes to represent the dynamic data and align the models for objective comparison.

Stackelberg games have been used in several deployed applications of game theory to make recommendations for allocating limited resources for protecting critical infrastructure. The resource allocation strategies are randomized to prevent a strategic attacker from using surveillance to learn and exploit patterns in the allocation. An important limitation of previous work on security games is that it typically assumes that attackers have perfect surveillance capabilities, and can learn the exact strategy of the defender. We introduce a new model that explicitly models the process of an attacker observing a sequence of resource allocation decisions and updating his beliefs about the defender's strategy. For this model we present computational techniques for updating the attacker's beliefs and computing optimal strategies for both the attacker and defender, given a specific number of observations. We provide multiple formulations for computing the defender's optimal strategy, including non-convex programming and a convex approximation. We also present an approximate method for computing the optimal length of time for the attacker to observe the defender's strategy before attacking. Finally, we present experimental results comparing the efficiency and runtime of our methods.