This project will develop a hybrid framework by integrating AI and machine learning methods with tabular information extraction and semantic modeling to improve the state-of-the-art precision and recall in tabular detection while maximizing the value of extracted information for industrial applications.
Advances in Whole Slide Imaging (WSI) and Machine Learning (ML) open new opportunities to create innovative solutions in healthcare and in particular digital pathology to increase efficiencies, reduce cost and most importantly improve patient care. This project envisions the creation of new automated ML tools including the design of a custom Convolution Neural Network (CNN) architecture for whole slide imaging in digital pathology. The custom CNN will be trained to learn different representations of histology tissues so that it can separate healthy from diseased tissues.
The purpose of this project is to investigate self-adaptive forecasting and anomaly prediction algorithms based on deep neural networks (DNNs). DNNs present a compelling technology due to their wide-spread availability through open-source projects (e.g. TensorFlow, MXNet). However, usability of DNNs in scenarios outside of image, speech or text pattern recognition is mostly unproven. This project aims to reduce the knowledge gap that exists in the usage of DNNs in the context of pattern recognition with DNNs in network management and network equipment manufacturing.
The integration of significant capacities of distributed energy resources (DERs) such as renewable wind and solar generation for a more sustainable energy future creates several challenges to the reliable and efficient operation of power distribution systems. These include: (i) Uncertain and intermittent nature of renewable generation compromises power quality for end-customers. (ii) Up-to-date distribution system network topologies are not well known and their real-time monitoring is limited. As a result, effective management of DERs is challenging.
In this project, a new method is developed to optimize the performance of an Unmanned Aerial Vehicle (UAV) for autonomous detection and on-the-job view-planning of infrastructure elements with the purpose of their accurate three-dimensional (3D) modeling. The existing view-planning approaches in the literature have mostly modeled non-complex or small-scale objects and have rarely been adapted to flying robots. In addition, the target object is often identified by human operators.
The research in this proposal examines the growing concern of frequency swings in modern power systems. With the increasing penetration of generation from renewable resources, the share of conventional modes of generation will be diluted and as a result the systemâs natural ability to maintain its frequency is diminished. Advanced converter systems may be able to help; however, their ability to do so is limited by several factors such as converter topology and device ratings, among other things.
In this research, a new approach to efficiently simulate large RLC represent power systems will be introduced and implemented in RTDS. The new approach utilize principle component analysis to search the subspace of state space vectors corresponding to a customer designed frequency band excitations and using projection method to form the reduced order system. Unlike other frequency dependent network equivalent methods, the proposed method reserves all internal information of original system and also inherently guarantees the passivity of the equivalent network.
Qubits are fundamental units for quantum computation. Photonics is a promising physical medium to realize large-scale quantum computation. One proposal to realize photonic qubits was proposed by Gottesman, Kitaev and Preskill (GKP). Here, the logical qubit is encoded into states of a bosonic mode or a quantum harmonic oscillator. It is expected that such a procedure will lead to a better quality and number of qubits.
In the real estate sector, a large volume of data is produced by businesses, commercial users and building visitors in a great variety of forms. For instance, three extensive sources of data come from unstructured text (e.g. documents, contracts), numerical data containing resources consumption and sensor/image-type data describing user behavior. A challenging problem for the sector is how to process the generated data into a useful asset that can provide insights to help business decisions, optimize user navigation and automate building-related processes.
This project applies wide-bandgap (WBG) transistors to voltage level multiplier module (VLMM) topology in motor inverter applications. It is expected that this approach can yield the benefits of WBG motor inverters (high motor efficiency, fast control response, lower motor torque ripple, close to ideal sinusoidal motor current waveform, smaller filter size, lower cost filter, etc.) while leveraging the benefits of VLMM (lower component cost, high frequency switching only at low voltage, filter-less output signal) to yield a commercially viable highly-efficient pure-sine motor inverter.