Probabilistic Deep Learning
Roberto discusses the innovative approach of using probabilistic numerics in deep learning, focusing on how to handle continuous functions instead of traditional vectors. He explains the development of a new convolutional layer based on partial differential equations, emphasizing the significance of translation equivariance and the application of generalized diffusion equations in neural networks. This method opens up exciting possibilities for model learning and performance.In this clip
From this podcast
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Probabilistic Numeric CNNs with Roberto Bondesan - #482
Related Questions