Vincent Cheung

University of Toronto, Toronto, Ontario
Thesis: Graphical Epitome Processing
Advisor: Brendan Frey

University of Manitoba, Winnipeg, Manitoba
GPA: 4.49/4.50
Governor General's Silver Medal for the highest undergraduate academic standing
Thesis: Signal Classification through Multifractal Analysis and Neural Networks
Advisor: Witold Kinsner

• Engineering principled machine learning solutions to a variety of problems, including those in computer vision and computational biology; solutions that both solve the problem at hand, yet are general enough to apply to completely different datasets or even other domains.
• Developing and using distributed computing systems in conjunction with machine learning algorithms.

University of Toronto, Dept. Elec. & Comp. Eng., Toronto, Ontario
Advisor: Brendan J. Frey

• Researched the modelling of a video sequence by a smaller sequence, called its "video epitome", with applications in object removal, video inpainting, frame interpolation, and video super-resolution. Implementations were done in Matlab, Java, and C#.
• Explored some of the open problems in the epitome framework including determining the size of the epitome and its computational issues.
• A distributed computing system was developed in Java, in which clients could be run on any machine that supports the Java JVM, including Windows, Mac, and Linux desktops as well as Linux clusters. This distributed representation resulted in a considerable speed improvement in both computing and utilizing video epitomes.

University of Toronto, Dept. Elec. & Comp. Eng., Toronto, Ontario
Advisor: Brendan J. Frey

• Collaborated with Charlie Boone (U. Toronto), Jack Greenblatt (U. Toronto), and Nevan Krogan (UCSF) to analyze their yeast gene interaction and cell biology datasets.
• Invented the "factorgram", which is a tool for visualizing multi-way associations in array data that overcomes some of the problems with clustergrams / dendrograms, where a given row or column may only belong in a single cluster.
• Pioneered "matrix tile analysis", the problem of decomposing a data matrix into a set of non-overlapping tiles, each of which is defined by a subset of rows and columns with no element belonging to more than one tile, in order to find multi-way associations in the data. This work was motivated by the desire to find protein complexes and pathways in gene interaction and gene expression data.

Google, Mountain View, California
Advisor: Ashutosh Garg

• Developed machine learning algorithms for performing computer vision tasks for mobile device applications.

Microsoft Research, Machine Learning and Applied Statistics, Redmond, Washington
Advisor: Nebojsa Jojic

• Developed machine learning algorithms for solving a variety of computer vision tasks, aiming for the algorithms to be as generally applicable as possible so that they were not tailored specifically to solve just one particular task or work in only very special cases.
• Modelling long distance correlations in patch based probabilistic graphical models was found to be important in manipulating the illumination of an image and in mimicking a physical walkthrough of an image. In both cases, the task was performed given only a single image and no knowledge of geometry.
• Sub-modular clustering was found to be useful for video segmentation.
• The use of video epitomes in performing biological cell tracking was examined.

University of Manitoba, Dept. Elec. & Comp. Eng., Winnipeg, Manitoba
Advisor: Witold Kinsner

• Researched the application of advanced computer engineering techniques to digital signal processing in order to analyze stochastic, non-stationary, self-similar signals such as electroencephalograms (EEGs), speech, and Internet traffic.
• Developed Java programs for neural networks, wavelets, fractal dimensions, distributed computing systems, genetic algorithms, data compression, and information entropy.
• Collaborated with Joseph Pear from the University of Manitoba's Psychology Department in analyzing spatio-temporal behavioural signals of Siamese fighting fish when presented with various stimuli during dishabituation experiments.
• The signals were found to be fractal in nature. Multifractal features were extracted from the signals and used by several neural networks, including probabilistic and complex domain neural networks, to perform classification of the fish behaviours.

1. F. J. Vizeacoumar, N. van Dyk, F. S. Vizeacoumar, V. Cheung, J. Li, Y. Sydorskyy, N. Case, Z. Li, A. Datti, C. Nislow, B. Raught, Z. Zhang, B. Frey, K. Bloom, C. Boone, and B. J. Andrews. Integrating high-throughput genetic interaction mapping and high-content screening to explore yeast spindle morphogenesis. The Journal of Cell Biology (JCB), Vol. 188, No. 1, January 2010.
2. V. Cheung, B. J. Frey, and N. Jojic. Video Epitomes. Intern. Journal of Computer Vision (IJCV), Vol. 76, No. 2, February 2008.
3. W. Kinsner, V. Cheung, K. Cannons, J. Pear, and T. Martin. Signal Classification through Multifractal Analysis and Complex Domain Neural Networks. IEEE Trans. on Systems, Man, and Cybernetics (SMC), Part C, Vol. 36, No. 2, March 2006.

4. V. Cheung, N. Jojic, D. Samaras. Capturing Long-Range Correlations with Patch Models. In Proc. IEEE Intern. Conf. Computer Vision and Pattern Recognition (CVPR), 2007.
5. I. E. Givoni, V. Cheung, and B. J. Frey. Matrix Tile Analysis. In Proc. Uncertainty in Artificial Intelligence (UAI), 2006.
6. V. Cheung, B. J. Frey, and N. Jojic. Video Epitomes. In Proc. IEEE Intern. Conf. Computer Vision and Pattern Recognition (CVPR), 2005. (Best paper honorable mention award)
7. W. Kinsner, V. Cheung, K. Cannons, J. Pear, and T. Martin. Signal Classification through Multifractal Analysis and Complex Domain Neural Networks. In Proc. IEEE Intern. Conf. Cognitive Informatics, August 2003.
8. V. Cheung, K. Cannons, W. Kinsner, and J. Pear. Signal Classification through Multifractal Analysis and Complex Domain Neural Networks. In Proc. IEEE Canadian Conf. Elec. and Comp. Eng., May 2003.

9. V. Cheung, I. E. Givoni, D. Dueck, and B. J. Frey. Factorgrams: A tool for visualizing multi-way associations in biological data. University of Toronto Technical Report PSI-2006-44, May 15, 2006.
10. V. Cheung. Efficiently Computing Distances in Patch-Based Models using the Shifted Cumulative Sum Algorithm. University of Toronto Technical Report PSI-2004-22, April 19, 2004.

11. V. Cheung, Graphical Epitome Processing. Ph.D. Thesis. Dept. of Electrical and Computer Engineering, University of Toronto, Toronto, ON, 2013, 90 pp.

12. V. Cheung and K. Cannons. Signal Classification through Multifractal Analysis and Neural Networks. BSc Thesis. Dept. of Electrical and Computer Engineering, University of Manitoba, Winnipeg, MB, 2003, 106 pp.

1. "Video Epitomes", IEEE Intern. Conf. Computer Vision and Pattern Recognition (CVPR), San Diego, June 22, 2005.
2. "Signal Classification through Multifractal Analysis and Complex Domain Neural Networks", IEEE Canadian Conf. Elec. and Comp. Eng., Montreal, May 5, 2003.

3. "Capturing Long-Range Correlations in Patch Models", Machine Learning Group Meeting, University of Toronto, Sept. 24, 2007.
4. "Video Epitomes"
   1. University of Windsor, Sept. 22, 2006.
   2. Google, Mountain View, California, June 21, 2006.
   3. University of Manitoba, May 13, 2005.
   4. The Learning Workshop, Snowbird, Utah, April 8, 2005.
5. "Modelling Motion Patterns with Video Epitomes", Machine Learning Group Meeting, University of Toronto, Oct. 18, 2004.
6. "Learning the 'Epitome' of a Video Sequence", Information Processing Workshop, University of Toronto, Aug. 11, 2004.
7. "An Introduction to Probabilistic Neural Networks", Signal and Data Compression Laboratory Group Meeting, University of Manitoba, June 10, 2002.

Shenzhen, China
Sponsored by: University of Toronto Association for Canada-China Exchange of Leadership

• Designed a course about Internet start-up companies and taught it to university students in China.

University of Toronto, Dept. Computer Science, Toronto, Ontario
Sponsored by: Canadian Institute for Advanced Research (CIAR)

• Presented prior work on epitomes as generative models and my work on video epitomes.

University of Toronto, Dept. Computer Science, Toronto, Ontario
Instructor: Richard Zemel

• Presented an introduction to Markov Random Fields and texture analysis.

MicroAge Technology Services, Winnipeg, Manitoba

• Gave tutorial sessions to customers on basic e-mail and web browsing usage. More in-depth lessons taught operability within Microsoft Office Word, Excel, and PowerPoint.

• Reviewer for IEEE Intern. Conf. on Comp. Vision (ICCV), IEEE Intern. Conf. on Comp. Vision and Pattern Recognition (CVPR), European Conf. on Comp. Vision (ECCV), Neural Information Processing Systems (NIPS), IEEE Trans. Pattern Analysis and Machine Intelligence (PAMI), Uncertainty in Artificial Intelligence (UAI), IEEE Trans. Systems, Man, and Cybernetics (SMC), and IEEE Trans. Circuits and Systems for Video Technology (TCSVT)
• Volunteer IEEE Canadian Conf. on Elec. & Comp. Eng., 2002, Winnipeg, MB
• Volunteer IEEE Canadian Conf. on Comp. & Soft. Eng. Education, 2002, Winnipeg, MB

Shape Collage Inc., Vancouver, British Columbia

• Algorithm and software development, website design, SEO, server administration, patent drafting, technology license agreement negotiations, marketing and PR, online advertising, customer support, project management, accounting, business strategy, business plan writing, pitching to VCs
• Bootstrapped to 6.5 million downloads and seven-figure revenues
• 2010 EO Graduate Student Entrepreneur Global Champion
• Winner of eight business plan competitions

VincentCheung.ca, Vancouver, British Columbia

• Wrote a program to download photos from Facebook and scrape the caption information for the jpeg IPTC metadata
• Developed a framework for encrypting and decrypting text for websites right in the browser using JavaScript

MicroAge Technology Services, Winnipeg, Manitoba

• Conducted troubleshooting and repaired hardware and software problems
• Assisted in the accounting and managerial areas

TVKMLMC Vending, Winnipeg, Manitoba

• Operated and maintained a vending machine and tracked weekly sales

Manitoba Telecom Services, Winnipeg, Manitoba

• Developed databases and designed web pages to access these databases

Manitoba Hydro, Winnipeg, Manitoba

• Serviced numerous internal clients by resolving their computer hardware, operating system, and software application problems

• Java, C++, C#, PHP, Javascript, Python, SQL, Matlab, LaTeX, Visual Basic, HTML, CSS, network and multi-threaded programming

• Volleyball, snowboarding, scuba diving, hiking, rollerblading, kayaking, ice skating, curling, running, table tennis, indoor rock climbing, traveling, cats

Available upon request