Macroeconomic Models for Performance and Investigation Prediction

Corporations are under a lot of scrutiny, especially when they annually release their financial reports to the government. If a corporation makes a mistake, or if an employee submits fraudulent information, or if it appears that either is the case, then they risk being asked to amend the filing by the government, which will cause their share price to suffer and force them to painstakingly redo the report at great expense. Caseware will sell software that can analyze these reports and determine if an amendment request is likely.

Assessing trust of artificial intelligence technology in the context of workplace relations

The goal of the project is to gain insight into individuals’ reactions to an artificial intelligence (AI) product currently in development at Kiite. The product is designed to fulfill some of the role responsibilities typically occupied by a manager. Trust is an important factor in both leader-employee relationships and in user experiences with AI-based systems. Thus, the partnership with Kiite offers a novel research opportunity to contribute to an emerging area of research on when and why humans are liable to (dis)trust AI technology in the workplace.

Content Delivery Networks to the Home

CONTENT Delivery Networks (CDNs) are large distributed infrastructures of replica servers placed in strategic locations. They deliver content to end-users with reduced latency by replicating content on surrogate servers. However they face a major challenge when content is delivered to end-users accessing a same content in home settings: inefficient bandwidth usage in the access network. There are as many streams from the replica server as end-users accessing the same content.

High-Fidelity Data Converters for Medical Diagnostics

Diagnostic medical devices work by translating our vital signs, such as neuron electrical activity and brain waves, into digital data that can be manipulated by a computer. High-speed computer processing improves diagnoses by presenting the physician with a numeric or graphical readout of important features extracted from the signal. Often, the ability of computer programs to extract the most diagnostically-relevant information is limited by how well the device can recognize and ignore background electrical noise common in clinical environments.

Research on Robust Face Recognition Algorithms

In recent years, face recognition algorithms based on deep neural networks have achieved human-level performance when tested on face recognition database. However, when put into real-world application, those algorithms are not robust enough, due to factors such as different lighting conditions, camera distance, and face orientations.

Dynamic Bandwidth Management

In the telecommunications industry, many schemas exist to cap or limit bandwidth at certain levels for clients. However, there are no real options available to allow clients to intelligently utilize spare bandwidth above their committed purchase rates. We propose to design, implement and evaluate novel bandwidth allocation mechanisms for high speed networks like the Cybera network. Cybera is a not-for-profit, technology-neutral organization responsible for driving Alberta’s economic growth through the use of digital technology.

Legal Question Answering with Machine Comprehension

ROSS Intelligence enables legal professionals to find analyze legal issues and find hidden information and cuts down on research time by using artificial intelligence specialized in legal research. Recent advances in neural networks applied to
natural language processing have brought results that are close to human performance in some tasks. However, this approach is still nascent in legal research and it has been identified as potentially fruitful.

Email Mining, Modeling, and Visualization

For this project, a data mining, visualization, and modeling technique will be developed and tested specifically for emails, using publicly available datasets. The mining will consist of gathering email and other potentially related datasets and cleaning those datasets. Cleaning will consist of removing duplicate or unnecessary information, as well as labeling data with basic information in order to ease training in the later steps. Next that data will be visualized in some form (graphs, charts, etc.) so that it may be more easily understood and a training model can be development.

Cited: Partnered Knowledge Mobilization Between Research and Media Organizations

Cited is a multimedia knowledge mobilization project that tells stories about research and academia to a general audience. It is experimenting with a unique co-creative approach that puts students, journalists, and researchers together on the same team. Mitacs interns will work with Cited media partners to conduct original research that builds interviews, documentaries, and other media related to research and academia—particularly in the social sciences and humanities. These will be distributed widely through a network of partners across North America.

Sparse Multivariate Polynomial Factorization

Factoring large polynomials is one of the main tools provided by mathematical software packages like Maple. It is used by scientists, engineers and mathematicians directly to simplify and study large formulas. It is also used inside Maple to do other tasks such as solving systems of polynomial equations. This project proposes to dramatically improve the speed of polynomial factorization so that larger polynomials can be factored and factored quickly, for example, in minutes instead of