# deep learning

Friday, October 18, 2019 - 9:05am - 9:50am

Mujdat Cetin (University of Rochester)

We present synthetic aperture radar (SAR) as a computational imaging modality, emphasizing aspects of radar that differentiate it from other imaging problems. In this context, we present samples of work resulting from two related lines of inquiry in our group: (1) sparsity-driven radar imaging, and (2) machine learning for radar imaging.

Wednesday, October 16, 2019 - 11:20am - 12:05pm

Jong Chul Ye (Korea Advanced Institute of Science and Technology (KAIST))

Encoder-decoder networks using convolutional neural network (CNN) architecture have been extensively used in deep learning approaches for inverse problems thanks to its excellent performance. However, it is still difficult to obtain coherent geometric view why such an architecture gives the desired performance.

Wednesday, October 16, 2019 - 9:50am - 10:35am

Yoram Bresler (University of Illinois at Urbana-Champaign)

A Generative Adversarial Network (GAN) trained to model the prior of images has been shown to perform better than sparsity-based regularizers in ill-posed inverse problems. We describe an approach along these lines, with some modifications and refinements, with the following features: (1) on a given class of images, it addresses different linear inverse problems without re-training the neural network; (2) it accelerates the computation substantially as compared to previous GAN-based methods; and (3) it comes with a recovery guarantee.

Wednesday, October 16, 2019 - 3:00pm - 3:45pm

Ge Wang (Rensselaer Polytechnic Institute)

Computer vision and image analysis are major application examples of deep learning. While computer vision and image analysis deal with existing images and produce features of these images (images to features), tomographic imaging produces images of multi-dimensional structures from experimentally measured “encoded” data as various tomographic features (integrals, harmonics, and so on, of underlying images) (features to images). Recently, deep learning is being actively developed worldwide for tomographic imaging, forming a new area of imaging research.

Wednesday, May 9, 2018 - 4:00pm - 4:30pm

Jiequn Han (Princeton University)

Developing algorithms for solving high-dimensional stochastic control problems and high-dimensional partial differential equations (PDEs) has been an exceedingly difficult task for a long time, due to the notorious difficulty known as the curse of dimensionality. In the first part of this talk, we develop a deep learning-based approach that directly solves high-dimensional stochastic control problems based on Monte-Carlo sampling.

Thursday, March 8, 2018 - 10:30am - 11:30am

Garrett Goh (Pacific Northwest National Laboratory)

With access to large datasets, deep neural networks (DNN) have achieved human-level accuracy in image and speech recognition tasks. However, in chemistry, data is inherently small and fragmented. In this work, we develop various approaches of using rule-based models and physics-based simulations to train ChemNet, a transferable and generalizable pre-trained network for small-molecule property prediction that learns in a weak-supervised manner from large unlabeled chemical databases.

Tuesday, March 6, 2018 - 9:00am - 10:00am

Herman van der Auweraer (Siemens)

The use of a priori digital models to build, evaluate and validate designs has become a standard practice in the product creation process. Furthermore, their use in production, operation and service has been increasing in the last years, for example in model predictive control, maintenance or assist systems.

Tuesday, May 17, 2016 - 2:00pm - 2:50pm

Shai Shalev-Shwartz (Hebrew University)

I will describe two contradicting lines of work. On one hand, a practical work on autonomous driving I was doing at Mobileye, in which deep learning is one of the key ingredients. On the other hand, theoretical work I was doing at the Hebrew university showing strong hardness of learning results. Bridging this gap is a great challenge. I will describe some approaches toward a solution, focusing on practically relevant theory and theoretically relevant practice.

Monday, February 22, 2016 - 1:15pm - 2:00pm

Stephen Wright (University of Wisconsin, Madison)

We survey some developments in machine learning and data analysis,

focusing on those in which optimization is an important

component. Some of these have possible relevance for industrial and

energy applications, for example, constraints and covariances could be

learned from process data rather than specified a priori. Some

possibilities along these lines will be proposed.

focusing on those in which optimization is an important

component. Some of these have possible relevance for industrial and

energy applications, for example, constraints and covariances could be

learned from process data rather than specified a priori. Some

possibilities along these lines will be proposed.

Wednesday, January 27, 2016 - 4:15pm - 5:10pm

René Vidal (Johns Hopkins University)

Matrix, tensor, and other factorization techniques are used in a wide range of applications and have enjoyed significant empirical success in many fields. However, common to a vast majority of these problems is the significant disadvantage that the associated optimization problems are typically non-convex due to a multilinear form or other convexity destroying transformation.