Regime Switch Analysis on Time-series Data for Financial Prediction

In recent years, the emergence of massive temporal data has become a reality in almost all aspects of social life, economic activity, security and defense, and poses a big challenge for existing methods. This project focuses on prediction from temporal data that arise ubiquitously in healthcare, social, industrial and financial fields. Events typically include changes in health status such as hospital readmission or death, evolution in social networks such that communities arise or vanish, modifications in energy consumption (e.g., wattage changes) and regime changes in stock markets.

Detection of Mental Health Conditions from Textual Device Communication

Research into child safety applications using Artificial Intelligence (AI) methods is a new area of investigation. SafeToNet is continuing to develop AI monitoring tools together with a team of researchers at the University of Ottawa. These tools, when used over time, will take advantage of outgoing text-based communications from devices to detect the early onset and progression of developmental and mental health issues in youth.

Texture Synthesis for Visual Effects: Improving Quality and Decreasing Computation Times

This proposal focuses on the automatic creation of color textures for 3D objects found in virtual content for movies, television, and advertisements. Such color details could correspond to the color variations seen at the surface of fabric or concrete. Another example of the problems we want to address consists of automatically creating color details for an animation of liquids such as mud. The solutions will reduce the computation times, increase the realism, and enable some methods to synthesize a broader variety of color textures.

Jordan Shapes for Deep Learning

The proposed project aims to develop a systematic approach for improving deep-learning-based computer vision systems by augmenting the local pixel data with the global shape data (more specifically, Jordan curves) and by adjusting system architectures to accommodate the augmented input. Three canonical computer vision problems will be investigated in this project. They are respectively image dehazing, alpha-matting, and face detection. The potential roles of Jordan curves in these applications will be examined.

Deep-learning-based Fine-grained Furniture Classification and Winning Strategy Recommendation

The project aims to develop a novel deep learning based computer vision system to identify different categories and sub-categories of the furniture and the associated attributes (such as color, shape, style, and material). It will also develop an automated recommendation system that can learn from the massive historical data and the on-going stream of data to adaptively adjust the parameter combination for each product to maximize the chance of winning the competition against other companies.

Research and improvement in label production process: waste reduction through adhesive quality improvement

As for any industry, the companies are constantly trying to improve their products. This can be done directly on the raw materials or on the production lines. But profitability can also be improved by reducing wastes while increasing production rates. So this project will investigate the processing conditions and adhesion formulations in terms of mechanical, thermal, chemical and adhesion properties. This will be done by optimization of both processing condition and concentration of the components.

Development of compact enhanced nebulization systems for inductively coupled plasma spectrometry

The measurement of toxins, such as arsenic, mercury, cadmium, lead and chromium for example, in food, beverages, environmental samples, waters, etc. must be carried out to verify that there is no danger. This requires analyses of numerous samples each day using instruments that can measure the small amounts that may be present.

Big Data Analysis and Management for Food Testing Industries Using Machine learning Technique

The project concerns big data analysis and management for food testing industries by using Machine learning technique. Any food product before and during its distribution to the market must go through an extensive safety and quality tests at food testing laboratories. During this process, various microbiology and analytical chemistry tests will be performed on food products to make sure about the safety and quality of the food products before offering them in the market.

A Machine Learning Approach for Digitalization of Engineering Specifications and Documents

Across industries, many engineering documents and drawings have accumulated over the past few decades. However, they are mostly archived in paper or rudimentary electronic form (typically in an image or PDF format), rendering information retrieval highly inconvenient. As such, a lot of valuable engineering data have been left unutilized or at very least, difficult to access. Unfortunately, the existing open source tools do not offer a simple remedy.

Optimization of group equivariant convolutional networks

The explosion of popularity of deep learning owes a lot to the success of convolutional neural networks, widely used in diverse fields including computer vision and natural language processing. Recently, the group equivariant convolutional neural network (G-CNN) was introduced, where equivariance of symmetries inherent in the data set is built in the architecture of the networks.