Uncategorized

Patients who have to undergo a magnetic resonance imaging (MRI) scan may be spared the ordeal of having to lie still in the scanner for up to 45 minutes, thanks to new compressive sensing technology developed in the groups of Rice ECE faculty Richard Baraniuk and Kevin Kelly.  The patented technology was recently licensed from Rice by Siemens Healthineers.

Magnetic resonance imaging (MRI) scanners equipped with compressive sensing operate much more quickly than current scanners.  Siemens Healthineers has applied the technology to help solve an important clinical problem: how to reduce long scan times while maintaining high diagnostic quality.  The result is the first clinical application of compressive sensing for cardiovascular imaging; it was approved for clinical use in February 2017 by the Food and Drug Administration. Thanks to compressive sensing, scans of the beating heart can be completed in as few as 25 seconds while the patient breathes freely. In contrast, in an MRI scanner equipped with conventional acceleration techniques, patients must lie still for six minutes or more and hold their breath as many as seven to 12 times throughout a cardiovascular-related procedure.

A. Mousavi, G. Dasarathy, R. G. Baraniuk, “DeepCodec: Adaptive Sensing and Recovery via Deep Convolutional Neural Networks,” arXiv:1707.03386, July 2017.

We develop a novel computational sensing framework for sensing and recovering structured signals called DeepCodec.  When trained on a set of representative signals, our framework learns to take undersampled measurements and recover signals from them using a deep convolutional neural network. In other words, it learns a transformation from the original signals to a near-optimal number of undersampled measurements and the inverse transformation from measurements to signals.  This is in contrast to conventional compressive sensing (CS) systems that use random linear measurements and convex optimization or iterative algorithms for signal recovery.  We compare our new framework with ℓ1-minimization from the phase transition point of view and demonstrate that it outperforms ℓ1-minimization in the regions of phase transition plot where ℓ1-minimization cannot recover the exact solution.  In addition, we experimentally demonstrate how learning measurements enhances recovery performance, speeds up training, and reduces the number of parameters to learn.

DeepCodec learns a transformation from signals x to measurement vectors y and an approximate inverse transformation from measurement vectors y to signals x using a deep convolutional network that consists of convolutional and sub-pixel convolution layers.

Recovery comparison of DeepCodec vs. LASSO (with optimal regularization parameter).

Rice University-based nonprofit OpenStax, which is already changing the economics of higher education by providing free textbooks to more than 1 million college students per year, today launched a low-cost, personalized learning system called OpenStax Tutor Beta that analyzes how students learn to offer them individualized homework and tutoring.  In development for three years, the system will be available this fall for three courses: college physics, biology and sociology.  While students study using OpenStax Tutor, it  learns how they learn — what they struggle with, what helps them most — and it uses that information to offer just-in-time remediation and enrichment.  The system provides personalized assessment and spaced practice, helping students focus their studying efforts on their weak areas and remember what they learned earlier in the course.

OpenStax Tutor Beta home
Read more
Inside Higher Ed story

Rice University professor and engineer Richard Baraniuk has been elected to the American Academy of Arts and Sciences. He is one of 228 new members announced today by the academy, which honors some of the world’s most accomplished scholars, scientists, writers, artists and civic, business and philanthropic leaders. The academy is one of the country’s oldest learned societies and independent policy research centers. It convenes leaders from the academic, business and government sectors to respond to the challenges facing - and opportunities available to - the nation and the world.

Rice University press release

Richard Baraniuk is one of 13 Vannevar Bush Faculty Fellows announced today by the U.S. Defense Department.  The fellows program provides extensive, long-term financial support for distinguished university scientists and engineers to pursue “blue sky” basic research that could produce revolutionary new technologies.  The program was launched in 2008 as the National Security Science and Engineering Faculty Fellowship (NSSEFF) program and renamed this year in honor of Vannevar Bush, the famed American engineer and inventor who headed U.S. scientific research during World War II and later helped found the National Science Foundation.  Baraniuk, Rice's Victor E. Cameron Professor of Electrical and Computer Engineering, is a leading expert on compressive sensing, a branch of signal processing that enables engineers to glean useful information from far fewer data samples than would typically be required.

DOD press release
Rice University press release

nailogoRichard Baraniuk, Rice University's Victor E. Cameron Professor of Electrical and Computer Engineering, has been elected a Fellow of the National Academy of Inventors.  He is one of 175 academic inventors named this year, and is now among 757 fellows representing 229 research universities and governmental and nonprofit research institutes.  NAI Fellows have demonstrated a prolific spirit of innovation to create or facilitate inventions that have made a tangible impact on quality of life, economic development and the welfare of society.  Combined, they are named inventors on more than 26,000 U.S. patents.

Press release

The National Educational Technology Plan for Higher Ed has been published by the US Department of Education.  The document is an outgrowth of the 2016 National Education Technology Plan (NETP), which presents a shared vision and call to action for transformational learning enabled by technology at all levels of our education system.  The NETP recommends actions that would enable everywhere, all-the-time learning and ensure greater equity and accessibility to learning opportunities over the course of a learner’s lifetime.  (Richard Baraniuk served on the Technical Working Group that reviewed drafts of the guide and provided feedback, writing, and examples from their experiences.)

semisupFinally, a deep convolutional network that doesn't require much "hand-holding" in the form of training examples.

Using the deep rendering mixture model, the network largely teaches itself how to distinguish handwritten digits using the MNIST dataset of 10,000 hand written digits.  Results presented this month at the Neural Information Processing Systems (NIPS) conference in Barcelona, Spain, demonstrate state-of-the-art training results using just 10 correct examples of each handwritten digit plus several thousand additional unlabelled examples.  The trained algorithm was more accurate at correctly distinguishing handwritten digits than almost all previous algorithms, even those trained with thousands of correct examples of each digit.

NIPS paper
More results
More information