Reinforcement learning defines the problem facing agents that learn to make good decisions through action and observation alone. To be effective problem solvers, such agents …
Humans are able to solve complex problems by distilling their knowledge of the world into simplified task-relevant representations and creating plans to achieve their goals. …
Just as an interconnected-computerized world has produced large amounts of data resulting in exciting challenges for machine learning, connected households with robots and smart devices …
Designing Artificial Intelligence (AI) is still reserved for experts, and the existing design paradigm follows a data-driven approach: domain experts start with a hypothetical model, …
Coreference Resolution is a fundamental natural language processing (NLP) problem, as it attempts to resolve which underlying discourse objects refer to one another. Further, it …
Recent natural disasters, such as Hurricane Katrina in 2005 and the Japan earthquake in 2011, have demonstrated that situational awareness, the focus of much research …
Solving combinatorial problems is an interplay between search and inference. In this thesis, we focus on search and investigate its important aspects. We start with …
Reinforcement learning (RL) techniques have led to remarkable results in challenging domains such as Atari games, Go, and Starcraft, suggesting that practical applications lie just …
Future collaborative robots must be capable of finding objects. As such a fundamental skill, we expect object search to eventually become an off-the-shelf capability for …
In this work, I present an implemented model that can learn interactively from natural language, enabling non-expert human trainers to convey complex tasks to machines …
Indoor scenes and the shapes that comprise them play a central role in our daily experiences. As virtual 3D representations of these environments become increasingly …
In order to intuitively and efficiently collaborate with humans, robots must learn to complete tasks specified using natural language. Natural language instructions can have many …
In this paper, we introduce a novel framework for improving intent classification performance in domains with limited labeled data. We call our framework RIPPED: Recursive …
Generally intelligent agents must learn and plan in complex environments, with sensors and actuators that support various behaviors and tasks. This complexity hinders decision making, …