Goto

Collaborating Authors

 Deep Learning


Approximate Inference Turns Deep Networks into Gaussian Processes

Neural Information Processing Systems

We present theoretical results aimed at connecting the training methods of deep learning and GP models. We show that the Gaussian posterior approximations for Bayesian DNNs, such as those obtained by Laplace approximation and variational inference (VI), are equivalent to posterior distributions ofGPregression models.



AI is indeed coming โ€“ but there is also evidence to allay investor fears

The Guardian

Traders on the floor of the New York Stock Exchange as markets fell on Friday morning trading after a steep drop on Thursday, as investors continue to worry about the impact of AI on business and the wider economy. Traders on the floor of the New York Stock Exchange as markets fell on Friday morning trading after a steep drop on Thursday, as investors continue to worry about the impact of AI on business and the wider economy. The message from investors to the software, wealth management, legal services and logistics industries this month has been clear: AI is coming for your business. The release of new, ever more powerful AI tools has coincided with a stock market slide, which has swept up sectors as diverse as drug distribution, commercial property and price comparison sites. Advances in the technology are giving increasing credulity to predictions that it could render millions of white-collar jobs obsolete - or, at least, eat into the profits of established companies.






Certified Robustness via Dynamic Margin Maximization and Improved Lipschitz Regularization

Neural Information Processing Systems

To improve the robustness of deep classifiers against adversarial perturbations, many approaches have been proposed, such as designing new architectures with better robustness properties (e.g., Lipschitz-capped networks), or modifying the