Nonlocal optimization of binary neural networks

no code implementations • 5 Apr 2022 • Amir Khoshaman, Giuseppe Castiglione, Christopher Srinivasa

We explore training Binary Neural Networks (BNNs) as a discrete variable inference problem over a factor graph.

A Path Towards Quantum Advantage in Training Deep Generative Models with Quantum Annealers

no code implementations • 4 Dec 2019 • Walter Vinci, Lorenzo Buffoni, Hossein Sadeghi, Amir Khoshaman, Evgeny Andriyash, Mohammad H. Amin

The hybrid structure of QVAE allows us to deploy current-generation quantum annealers in QCH generative models to achieve competitive performance on datasets such as MNIST.

Quantum Variational Autoencoder

no code implementations • 15 Feb 2018 • Amir Khoshaman, Walter Vinci, Brandon Denis, Evgeny Andriyash, Hossein Sadeghi, Mohammad H. Amin

We show that our model can be trained end-to-end by maximizing a well-defined loss-function: a 'quantum' lower-bound to a variational approximation of the log-likelihood.

DVAE++: Discrete Variational Autoencoders with Overlapping Transformations

no code implementations • ICML 2018 • Arash Vahdat, William G. Macready, Zhengbing Bian, Amir Khoshaman, Evgeny Andriyash

Training of discrete latent variable models remains challenging because passing gradient information through discrete units is difficult.

Image Generation