Neural Processes
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Elre Oldewage.
Neural Processes (NPs) are a recently proposed method for using meta-learning to train neural networks to predict stochastic processes. They can be applied in regression tasks that require prediction with uncertainty in the small-data regime and fast test-time inference. Furthermore, the meta-learning framework allows NPs to learn intricate structure in the stochastic process directly from the data, allowing them to be applied to image data. In this reading group, we will introduce the basic NP architecture and go through some of the many kinds of NP that have been proposed since 2018, including the attentive NP and convolutional NP.
Suggested reading:
The talk will largely follow the content of this blog, which also includes code/pre-trained models:
https://yanndubs.github.io/Neural-Process-Family
Conditional Neural Processes: https://arxiv.org/abs/1807.01613
Neural Processes: https://arxiv.org/abs/1807.01622
This talk is part of the Machine Learning Reading Group @ CUED series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|