The Rise of Neural Priors

Explore the concept of neural priors in AI and machine learning through this insightful 28-minute talk by Prof. Simon Lucey, Director of the Australian Institute for Machine Learning. Delve into a new direction that challenges the conventional reliance on massive training datasets and addresses issues of data mismatch between train and test sets. Discover how neural priors leverage the implicit regularization properties of neural network architectures to tackle problems with limited training data or out-of-distribution bias. Learn about practical applications of neural priors in augmented reality, autonomous driving, and noisy signal recovery, and understand their potential impact on industry. Gain valuable insights from Prof. Lucey's extensive experience in computer vision, machine learning, and robotics as he draws inspiration from past AI researchers to unlock computational and mathematical models underlying visual perception processes.