Citadel Chief AI Officer Li Deng has been named a Fellow of the Canadian Academy of Engineering (CAE) in recognition of his notable achievements in deep learning and speech recognition.
An accomplished AI scientist, Deng joined the US$30 billion hedge fund Citadel as its Chief AI Officer in May 2017. Prior to that he was Chief Scientist of AI at Microsoft, where he led the company’s AI school and founded the Deep Learning Technology Center. Deng has also served as an Affiliate Full Professor at the University of Washington since 2000.
Deng obtained his Master’s and a PhD in Electrical and Computer Engineering at the University of Wisconsin-Madison. His main research focuses include AI, machine learning, signal processing, financial engineering, speech recognition, and natural language processing. In 1989 he joined the University of Waterloo in Canada, where he was an assistant professor (1989-1992), tenured associate professor (1992-1996), and tenured full professor (1996-2000). In 2000 he joined Microsoft, where he focused on speech recognition, natural language understanding, and signal processing research using Bayesian and other machine learning approaches to deep generative modeling.
In 2006, “Godfather of Deep Learning” Dr. Geoffrey Hinton published the paper A fast learning algorithm for deep belief nets, which showed how a deep belief network with many hidden layers could create a well-behaved generative model representing the joint distribution of handwritten digit images and their labels. The results inspired Deng to apply deep neural networks to speech recognition.
Some of Deng’s most notable contributions are his pioneering work, from 2009 to 2013, that ushered in the wide adoption of large-scale deep learning techniques in speech recognition.
At NIPS 2009 Deng and Hinton co-organized the workshop Deep Learning for Speech Recognition and Related Applications, where they demonstrated that deep architectures trained with novel approaches could outperform previous methods on a variety of small tasks of speech recognition tasks.
Hinton and Deng published and reviewed theirs and other organizations’ results in the 2012 paper Deep Neural Networks for Acoustic Modeling in Speech Recognition, which has received more than 5700 citations as of April 2019. The paper proposed and elaborated on deep feed-forward neural networks as an alternative for acoustic modeling in speech recognition that “takes several frames of coefﬁcients as input and produces posterior probabilities over HMM states as output.”
“Deep learning is revolutionary in the sense that it is disrupting all previous paradigms by allowing the most effective representation of speech signals … in the joint temporal and spatial domain,” says Deng.
In 2015 Deng was awarded the IEEE Signal Processing Society Technical Achievement Award for his Outstanding Contributions to Deep Learning and Automatic Speech Recognition.
In its citation letter of April 2019, the CAE recognized Deng’s “original and landmark research, over 30 years, combined with his outstanding leadership in advancing engineering knowledge, have culminated in the spectacular effects of deep learning and AI on society today.”
Journalist: Tony Peng | Editor: Michael Sarazen