Total: 19
Cross-lingual Learning can help to bring state-of-the-art deep learning solutions to smaller languages. These languages in general lack resource for training advanced neural networks. With transfer of knowledge across languages we can improve the results for various NLP tasks.
As people age, it is critical that they maintain not only their physical health, but also their cognitive health―for instance, by engaging in cognitive exercise. Recent advancements in AI have uncovered novel ways through which to facilitate such exercise. In this thesis, I propose the first human-robot dialogue system designed specifically to promote cognitive exercise in elderly adults, through discussions about interesting metaphors in books. I describe my work to date, including the development of a new, large corpus and an approach for automatically scoring metaphor novelty. Finally, I outline my plans for incorporating this work into the proposed system.
Industries are on the brink of widely accepting a new paradigm for organizing production by having autonomous robots manage in-factory processes. This transition from static process chains towards more automation and autonomy poses new challenges in terms of, e.g., efficiency of production processes. The RoboCup Logistics League (RCLL) has been proposed as a realistic testbed to study the above mentioned problem at a manageable scale. In RCLL, teams of robots manage and optimize the material flow according to dynamic orders in a simplified factory environment. In particular, robots have to transport workpieces among several machines scattered around the factory shop floor. Each machine performs a specific processing step, orders that denote the products which must be assembled with these operations are posted at run-time and require quick planning and scheduling. Orders also come with a delivery time window, therefore introducing a temporal component into the problem. Though there exist successful heuristic approaches to solve the underlying planning and scheduling problems, a disadvantage of these methods is that they provide no guarantees about the quality of the solution. A promising solution to this problem is offered by the recently emerging field of Optimization Modulo Theories (OMT), where Satisfiability Modulo Theories (SMT) solving is extended with optimization functionalities. In this paper, we present an approach that combines bounded model checking and optimization to generate optimal controllers for multi-robot systems. In particular, using the RoboCup Logistics League as a testbed, we build formal models for robot motions, production processes, and for order schedules, deadlines and rewards. We then encode the synthesis problem as a linear mixed-integer problem and employ Optimization Modulo Theories to synthesize controllers with optimality guarantees.
Combinatorial optimisation problems often contain uncertainty that has to be taken into account to produce realistic solutions. A common way to describe the uncertainty is by means of scenarios, where each scenario describes different potential sets of problem parameters based on random distributions or historical data. While efficient algorithmic techniques exist for specific problem classes such as linear programs, there are very few approaches that can handle general Constraint Programming formulations subject to uncertainty. The goal of my PhD is to develop generic methods for solving stochastic combinatorial optimisation problems formulated in a Constraint Programming framework.
Fine-grained Entity Recognition (FgER) is the task of detecting and classifying entity mentions into more than 100 types. The type set can span various domains including biomedical (e.g., disease, gene), sport (e.g., sports event, sports player), religion and mythology (e.g., religion, god) and entertainment (e.g., movies, music). Most of the existing literature for Entity Recognition (ER) focuses on coarse-grained entity recognition (CgER), i.e., recognition of entities belonging to few types such as person, location and organization. In the past two decades, several manually annotated datasets spanning different genre of texts were created to facilitate the development and evaluation of CgER systems (Nadeau and Sekine 2007). The state-of-the-art CgER systems use supervised statistical learning models trained on manually annotated datasets (Ma and Hovy 2016). In contrast, FgER systems are yet to match the performance level of CgER systems. There are two major challenges associated with failure of FgER systems. First, manually annotating a large-scale multi-genre training data for FgER task is expensive, time-consuming and error-prone. Note that, a human-annotator will have to choose a subset of types from a large set of types and types for the same entity might differ in sentences based on the contextual information. Second, supervised statistical learning models when trained on automatically generated noisy training data fits to noise, impacting the model’s performance. The objective of my thesis is to create a FgER system by exploring an off the beaten path which can eliminate the need for manually annotating large-scale multi-genre training dataset. The path includes: (1) automatically generating a large-scale single-genre training dataset, (2) noise-aware learning models that learn better in noisy datasets, and (3) use of knowledge transfer approaches to adapt FgER system to different genres of text.
Currently, much of machine learning is opaque, just like a "black box." However, in order for humans to understand, trust and effectively manage the emerging AI systems, an AI needs to be able to explain its decisions and conclusions. In this paper, I propose an argumentation-based approach to explainable AI, which has the potential to generate more comprehensive explanations than existing approaches.
We consider single-agent, two-agent turn-taking and cooperative security games with Inverse Geodesic Length (IGL) as utility metric. We focus on the single-agent vertex deletion problem corresponding to IGL called MinIGL. Specifically, given a graph G, a budget k and a target inverse geodesic length T, does there exist a subset of vertices S of size k such that by deleting S the graph induced on the remaining vertices in G has IGL at most T. We cite our recently published work to report the results on the computational and parameterized complexity of MinIGL. Furthermore, we briefly state the problems we are interested to study in future.
Worldwide, conservation agencies employ rangers to protect conservation areas from poachers. However, agencies lack the manpower to have rangers effectively patrol these vast areas frequently. While past work has modeled poachers’ behavior so as to aid rangers in planning future patrols, those models’ predictions were not validated by extensive field tests. In my thesis, I present a spatio-temporal model that predicts poaching threat levels and results from a five-month field test in Uganda’s Queen Elizabeth Protected Area (QEPA). To my knowledge, this is the first time that a predictive model has been evaluated through such an extensive field test in this domain. These field test will be extended to another park in Uganda, Murchison Fall Protected Area, shortly. Main goals of my thesis are to develop the best performing model in terms of speed and accuracy and use such model to generate efficient and feasible patrol routes for the park rangers.
Graphical models provide a powerful framework for reasoning under uncertainty, and an influence diagram (ID) is a graphical model of a sequential decision problem that maximizes the total expected utility of a non-forgetting agent. Relaxing the regular modeling assumptions, an ID can be flexibly extended to general decision scenarios involving a limited memory agent or multi-agents. The approach of probabilistic planning with IDs is expected to gain computational leverage by exploiting the local structure as well as representation flexibility of influence diagram frameworks. My research focuses on graphical model inference for IDs and its application to probabilistic planning, targeting online MDP/POMDP planning as testbeds in the evaluation.
I present the outline of my dissertation work, Identifying Private Content for Online Image Sharing. Particularly, in my dissertation, I explore learning models to predict appropriate binary privacy settings (i.e., private, public) for images, before they are shared online. Specifically, I investigate textual features (user-annotated tags and automatically derived tags), and visual semantic features that are transferred from various layers of deep Convolutional Neural Network (CNN). Experimental results show that the learning models based on the proposed features outperform strong baseline models for this task on the Flickr dataset of thousands of images.
Past work in security games has mainly focused on the problem static resource allocation; how to optimally deploy a given fixed team of resources. My research aims to address the challenge of integrating operational planning into security games, where resources are heterogeneous and the defender is tasked with optimizing over both the investment into these resources, as well as their deployment in the field. This allows the defender to design more adaptive strategies, reason about the efficiency of their use of these resources as well as their effectiveness in their deployment. This thesis explores the challenges in integrating these two optimization problems in both the single stage and multi-stage setting and provides a formal model of this problem, which we refer to as the Simultaneous Optimization of Resource Teams and Tactics (SORT) as a new fundamental research problem in security games that combines strategic and tactical decision making. The main contributions of this work are solution methods to the SORT problem under various settings as well as exploring various types of tradeoffs that can arise in these settings. These include managing budget for investment in resources as well as capacity constraints on use of resources. My work addresses scenarios when the tactical decision problem (optimal deployment) is difficult, and thus evaluating the performance of any given team is difficult. Additionally, I address domains where we are tasked with making repeated strategic level decision and where, due to changing domain features, fluctuations in time dependent processes or the realization of uncertain parameters in the problem, it becomes necessary to re-evaluate and adapt to new information.
My main research motivation is to develop complete autonomous agents that interact with people socially. For an agent to be social with respect to humans, it needs to be able to parse and process the multitude of aspects that comprise the human cultural experience. That in itself gives rise to many fascinating learning problems. I am interested in tackling these fundamental problems from an empirical as well as a theoretical perspective. Music, as a general target domain, serves as an excellent testbed for these research ideas. Musical skills---playing music (alone or in a group), analyzing music or composing it---all involve extremely advanced knowledge representation and problem solving tools. Creating "musical agents"---agents that can interact richly with people in the music domain---is a challenge that holds the potential of advancing social agents research, and contributing important and broadly applicable AI knowledge. This belief is fueled not just by my background in computer science and artificial intelligence, but also by my deep passion for music as well as my extensive musical training. One key aspect of musical intelligence which hasn’t been sufficiently studied is that of sequential decision-making. My thesis strives to answer the following question: How can a sequential decision making perspective guide us in the creation of better music agents, and social agents in general? More specifically, this thesis focuses on two aspects of musical intelligence: music recommendation and multiagent interaction in the context of music.
Autonomous vehicles are quickly becoming an important part of human society for transportation, monitoring, agriculture, and other applications. In these applications, there is a fundamental tradeoff between safety and efficiency that is especially salient when the autonomous vehicles interact directly with humans. A key to maintaining safety without sacrificing efficiency is dealing with uncertainty properly so that robots can be assertive when it is appropriate and careful in dangerous situations. The research that will be presented in my thesis uses the partially observable Markov decision process framework to approach this challenge, exploring several applications and proposing a new solution approach that is able to handle continuous action and observation spaces, a qualitative improvement over current methods.
The last two decades have seen extraordinary advances in industrial applications of constraint satisfaction techniques, while combinatorial problems have been pushed to the sidelines. We propose a comprehensive analysis of the state of the art in constraint satisfaction problems when applied to combinatorial problems in areas such as graph theory, set theory, algebra, among others. We believe such a study will provide us with a deeper understanding about the limitations we still face in constraint satisfaction problems.
Machine learning applications to electrical time series data will have wide-ranging impacts in the near future. Electricity disaggregation holds the promise of reducing billions of dollars of electrical waste every year. In the power grid, automatic classification of disturbance events detected by phasor measurement units could prevent cascading blackouts before they occur. Additional applications include better market segmentation by utility companies, improved design of appliances, and reliable incorporation of renewable energy resources into the power grid. However, existing machine learning methods remain unimplemented in the real world because of limiting assumptions that hinder performance. My research contributions are summarized as follows: In electricity disaggregation, I introduced the first label correction approach for supervised training samples. For unsupervised disaggregation, I introduced event detection that does not require parameter tuning and appliance discovery that makes no assumptions on appliance types. These improvements produce better accuracy, faster computation, and more scalability than any previously introduced method and can be to applied natural gas disaggregation, water disaggregation, and other source separation domains. My current work challenges long-held assumptions in time series shapelets, a classification tool with applicability in electrical time series and dozens of additional domains.
My research addresses the problem faced by a defender who must screen objects for potential threats that are coming into a secure area. The particular domain of interest for my work is the protection of cyber networks from intrusions given the presence of a strategic adversary. My thesis work allows fora defender to use game-theoretical methods that randomize her protection strategy and introduces uncertainty to the adversary that makes it more difficult to attack the defender’s network successfully.
As we move towards autonomous machines responsible for making decisions previously entrusted to humans, there is an immediate need for machines to be able to explain their behavior and defend the reasonableness of their actions. To implement this vision, each part of a machine should be aware of the behavior of the other parts that they cooperate with. Each part must be able to explain the observed behavior of those neighbors in the context of the shared goal for the local community. If such an explanation cannot be made, it is evidence that either a part has failed (or was subverted) or the communication has failed. The development of reasonableness monitors is work towards generalizing that vision, with the intention of developing a system-construction methodology that enhances both robustness and security, at runtime (not static compile time), by dynamic checking and explaining of the behaviors of parts and subsystems for reasonableness in context.
We present a new sampling scheme for approximating hard to compute queries over graphical models, such as computing the partition function. The scheme builds upon exact algorithms that traverse a weighted directed state-space graph representing a global function over a graphical model (e.g., probability distribution). With the aid of an abstraction function and randomization, the state space can be compacted (trimmed) to facilitate tractable computation, yielding a Monte Carlo estimate that is unbiased. We present the general idea and analyze its properties analytically and empirically.
With the advent of personal assistants such as Siri and Alexa, there has been a renewed focus on dialog systems, specifically open domain conversational agents. Dialog is a challenging problem since it spans multiple conversational turns. To further complicate the problem, there are many contextual cues and valid possible utterances. Dialog is fundamentally a multiscale process given that context is carried from previous utterances in the conversation; however, current neural methods lack the ability to carry human-like conversation. Neural dialog models are based on recurrent neural network Encoder-Decoder sequence-to-sequence models (Sutskever, Vinyals, and Le, 2014; Bahdanau, Cho, and Bengio, 2015). However, these models lack the ability to create temporal and stylistic coherence in conversations. We propose to incorporate dialog acts (such as Statement-non-opinion ["Me, I'm in the legal department."], Acknowledge ["Uh-huh."]) and discourse connectives (e.g. "because," "then"), utterance clustering and domain prediction, and style shifting using hierarchical methods. In particular, we show that clustering of utterance representations automatically allows for a unified hierarchical approach to discourse, domain, and style.