Sageev Oore completed an undergraduate degree in Mathematics (Dalhousie), and MSc and PhD degrees in Computer Science (University of Toronto) working with Geoffrey Hinton. In his MSc, he developed a minimally-supervised learning algorithm for robot localization– a precursor of particle filtering and SLAM; and in his PhD, co-supervised also by Demetri Terzopoulos, he developed a real-time tool for generating character animation. A professional musician, Sageev has performed as soloist with orchestras both as a classical pianist and as a jazz improviser. Together with his brother Dani, he recorded an album combining classical art songs with improvisation. In 2016, Sageev’s long-standing fascination with combining machine learning & music surpassed his long-standing resistance to that same topic, and he joined the Magenta project at Google Brain (Mountain View, California) as a visiting scientist, applying deep learning approaches to music. Having now joined the Faculty of Computer Science at Dalhousie University and the Vector Institute, he is building a research programme exploring machine learning in computational creativity.
- Machine learning to generate music, audio, text and images
- Computational creativity: what it might be (exploring its limits) and how to work with it (e.g. developing tools for artists)
- With Ian Simon: Performance RNN, a system for generating music with expressive timing & dynamics https://magenta.tensorflow.org/performance-rnn
Estimating Severity of Depression From Acoustic Features and Embeddings of Natural Speech
High Frequency-Low Amplitude Oscillometry: Continuous Unobtrusive Monitoring of Respiratory Function on PAP Machines
Controlling BigGAN Image Generation with a Segmentation Network
Significance of Speaker Embeddings and Temporal Context for Depression Detection