Automatic Approach to Design Efficient Deep Neural Networks

Deep neural networks have demonstrated state-of-the-art modeling accuracy on a wide range of real-life problems, with some cases surpassing human performance. Despite the promise of deep neural networks as an enabling technology for a large number of industries and fields, there are two particular key challenges in the design of deep neural networks in real-world, operational scenarios. First, the design of deep neural networks is a very time consuming process for a machine learning expert, and often results in complex, non-optimal deep neural networks.

Development of a reliable and scalable underwater acoustic modem for networked applications

The proposed project represents a critical effort towards developing the enabling communication technology for the future of subsea connectivity where conventional communications technologies such as Wi-Fi and GPS cannot be used. The intern will work to completely overhaul traditional underwater communications methodologies and advance acoustic communications towards the higher reliability and data rates needed for future underwater networked applications and deployments.

Visual Analytic Tool for Lessons Learned Retrieval and Decision Making

According to the World Petroleum Council (WPC), the average age of employees in Oil and Gas companies is 50 years, and it is estimated that in the next 5 years 40-60% of them will retire. One consequence of this age-related crisis is losing the accumulated knowledge by retiring “gray-beards”. In this scenario, new software technologies are mandatory to retain decades of expertise and transfer it to new employees.

Surface testing: A solution for surface degradation of industrial materials in harsh environment

This project will demonstrate the cause and effect of industrial material failures in the harsh Newfoundland climate under the general umbrella of corrosion. Our main focus is materials used in the marine and oil & gas industries.

Assessment of deep learning for analyzing radar signals in maritime environment

The proposed internships aim at investigating the relevance of deep learning (DL) techniques for target detection in radar data processing. More specifically, we are looking to demonstrate the feasibility of DL techniques to deal with unusual types of data (i.e., radar data) in situations where an well performing processing with classical techniques is a challenge (e.g., detection of objects in noisy scenes from a maritime environment caused by the interference produced by the reflection of the radar waves on the sea).

Leverage on Artificial Intelligence in Capacity Management: Predict IT assets usage based on Business events

The Societe Generale Bank possesses a network of trading applications which generate the hardware consumption data (CPU, memory and network communication).

Motion fields with deep reinforcement learning for real-time character animation

Character motion in games and animations often have high requirements of realism, aesthetics, and interactivity. For instance, in soccer simulation games, users control the players to move in different directions and perform actions such as passing and shooting. Modern data-driven approaches like motion fields provide convenient ways to synthesizing natural motions from a given database of motion capture data. In this work, we look to improve motion fields by leveraging deep reinforcement learning.

Translational Research in User Experience Design for Personalized Health

Personalized health is increasingly gaining public attention in the media as the future of healthcare. Personalized health is the idea that medical treatment will be tailored to the individual based on their predicted response or risks of disease.

RESPOND (Resource Efficient Smart Packet Optical Network Design): A Novel Packet-Optical Design and Optimization Framework for Next Generation Networks

The focus of the project is to develop an packet-optical network resource optimization model that minimizes the total network cost across IP-optical platform while meeting the following requirements: (i) Offers full protection from any network node and link level failure. (ii) Ability to handle large scale networks and traffic demand (i.e., network scalability). (iii) Meets end-to-end latency requirement. (iv) Provides efficient link utilization across the packet-optical networks. (v) Ability to forecast network capacity augment requirement.

RBC-Toronto Quantum Key Distribution Network Development

The interns will develop a quantum communication network built on RBC’s optical fiber infrastructure and perform secure commercial transactions using quantum-generated secure keys in the integrated classical communication network. The quantum communication network will be based on the measurement-device-independent quantum key distribution technology, which is developed by Prof. Hoi-Kwong Lo’s and Prof. Li Qian’s groups at the University of Toronto. It will be an important milestone towards cybersecurity in the financial sector in Canada.