Sorry, you need to enable JavaScript to visit this website.

Active learning is a form of machine learning which combines supervised learning and feedback to minimize the training set size, subject to low generalization errors. Since direct optimization of the generalization error is difficult, many heuristics have been developed which lack a firm theoretical foundation. In this paper, a new information theoretic criterion is proposed based on a minimax log-loss regret formulation of the active learning problem. In the first part of this paper, a Redundancy Capacity theorem for active learning is derived along with an optimal learner.

We focus on a wireless sensor network powered with an energy beacon, where sensors send their measurements to the sink using the harvested energy. The aim of the system is to estimate an unknown signal over the area of interest as accurately as possible. We investigate optimal energy beamforming at the energy beacon and optimal transmit power allocation at the sensors under non-linear energy harvesting models. We use a deep reinforcement learning (RL) based approach where multi-layer neural networks are utilized.

Video event recognition based on audio and visual modalities is an open research problem. The mainstream literature on video event recognition focuses on the visual modality and does not take into account the relevant information present in the audio modality. We propose to study several fusion architectures for the audio-visual recognition task of video events. We first build classical fusion architectures using concatenation, addition or Multimodal Compact Bilinear pooling (MCB).

Building floor plans with locations of safety, security, and energy assets such as IoT sensors, fire alarms, etc. are vital for climate control, emergency response, safety, and maintenance of building infrastructure. Existing approaches to building survey are tedious, error prone, and involve an operator with a clipboard and pen, enumerating and localizing assets in each room. We propose an interactive method for a human operator to use an app on a smartphone, which can detect, classify, and localize assets of interest, to expedite such a task.

The supervised learning paradigm is limited by the cost - and sometimes the impracticality - of data collection and labeling in multiple domains. Self-supervised learning, a paradigm which exploits the structure of unlabeled data to create learning problems that can be solved with standard supervised approaches, has shown great promise as a pretraining or feature learning approach in fields like computer vision and time series processing. In this work, we present self-supervision strategies that can be used to learn informative representations from multivariate time series.