My name is Sam Cantor. I am a software developer intern at Kings Distributed Systems, working on the Distributed Compute Protocol for Distributed Compute Labs. I participate in the planning, development, and testing of new software, which includes documentation, APIs, and overall system architecture. Part of my work at DCP includes researching new and innovative ways to distribute neural network training. As datasets and machine learning models increase in size and complexity, access to computing power to train these models becomes vital. The ability to use DCP’s platform for machine learning would drastically reduce the amount of computing time for training. The goal of my individual research at DCP is to establish a system for distributing machine/deep learning tasks on the platform, with a focus on developing a program to train convolutional neural networks on the platform for image/video processing.
I am currently in my third year at Queen's University, studying computer engineering. I have 5+ years of programming experience and I specialize in machine learning. My most recent project involved exploring the use of Generative Adversarial Networks (GANs) in music generation. The network was trained from data collected using the Spotify API, and compared music clips to their corresponding cover artwork. The trained model can now generate a new song when given an input image.
Outside of work hours, I enjoy spending my time performing and producing music. I have studied classical, rock and jazz piano, and had the opportunity and the privilege to be taught from notable musicians such as Phil X, Dave Limina, and Jordan Rudess, among others. I love jamming and sharing my passion for music with others.
- Parallel CNNs — An overview of my research at DCL. Includes a low-level description of how CNNs work and possible methods to distribute training
- schedmsg — A system for interacting with the workers connected to DCP to display broadcast messages or prompt updates