"It describes everything you see outside of your window," said Jiali Wang, an environmental scientist at the U.S. Department of Energy's (DOE) Argonne National Laboratory, "from the clouds, to the sun's radiation, to snow to vegetation -- even the way skyscrapers disrupt the wind." The myriad characteristics and causes of weather and climate are coupled together, communicating with one another. Scientists have yet to fully describe these complex relationships with simple, unified equations. Instead, they approximate the equations using a method called parameterization in which they model the relationships at a scale greater than that of the actual phenomena. Although parameterizations simplify the physics in a way that allows the models to produce relatively accurate results in a reasonable time, they are still computationally expensive.

At RSNA 2019, attendees will have the opportunity to learn how artificial intelligence (AI) is transforming women's imaging, such as enhancing breast imaging, improving radiologists' workflow, and reducing recall rates with digital mammography. This year's conference will also feature presentations on shear-wave elastography, contrast-enhanced spectral mammography, and more. Presentations will highlight how AI can improve or enhance breast imaging, covering topics such as the use of deep learning to reduce digital breast tomosynthesis (DBT) reading time and even further improve its cancer detection capability; the benefits of allowing AI algorithms to sift through mammograms and eliminate low-malignancy exams, thus improving radiologists' workflow; using AI as a tool to reduce the recall rate on digital mammography; and machine learning-based evaluation of DBT screening through the creation of customized synthesized 2D images. In fact, the RSNA plans to kick off the week with a Deep Learning Classroom session that's part of its AI Showcase, and the session will be repeated throughout the meeting. Yet even with all this interest in AI, RSNA 2019 will offer attendees a chance to explore a variety of other women's imaging matters as well. Keep an eye out for presentations on the feasibility of supine MRI-navigated ultrasound in breast cancer patients, the effect of computer-aided diagnosis on breast ultrasound interpretation, using shear-wave elastography to predict breast cancer patient response to chemotherapy, and contrast-enhanced spectral mammography for diagnostic workup of BI-RADS 4 lesions on breast MRI, to name a few.

While advances in hybrid imaging and other cutting-edge modalities seem to get all the attention, digital radiography (DR) maintains its solid foundation of diagnostic support for imaging centers, emergency departments, outpatient clinics, and mobile operations worldwide. Based on the myriad of presentations scheduled for RSNA 2019, DR arguably might be the modality that will benefit most from artificial intelligence (AI) and deep-learning algorithms. As DR-specific applications are tested and validated, radiologists could soon turn to AI to generate clinically relevant x-ray reports, diagnose fractures and other ailments throughout the body, visualize motion, and increase the accuracy of their readings. AI also is expected to significantly reduce the time and effort it takes for some of the tasks that are necessary but tedious and time-consuming with digital radiography, such as double reading and confirming normal results. Beyond AI, researchers continue to explore hardware improvements in DR systems and develop new technologies to sharpen image quality and shorten exam and procedure times.

This is an episode from Josh Reppun's "What School Could Be in Hawai'i," a podcast on the people, technology and methodologies pushing the mantle of education in the 50th state. Susannah Johnson is the founder of Individualized Realized, an education consultancy aimed at meeting educators where they are – as she did in the classroom with students for thirteen years – on the path to student-centered, authentic, globally minded, and liberated learning. In the move towards student-centered learning technology is essential for individualized learning. Over ten years developing a fully individualized program, the use of technology not only opens up learning to be multidimensional, but also for the asynchronous management of dozens of curricula. When students own their own learning, technology moves beyond learning tool to become a partner for that learning.

Radiologists assisted by deep-learning based software were better able to detect malignant lung cancers on chest X-rays, according to research published in the journal Radiology. "The average sensitivity of radiologists was improved by 5.2% when they re-reviewed X-rays with the deep-learning software," said Byoung Wook Choi, M.D., Ph.D., professor at Yonsei University College of Medicine, and cardiothoracic radiologist in the Department of Radiology in the Yonsei University Health System in Seoul, Korea. "At the same time, the number of false-positive findings per image was reduced." Dr. Choi said the characteristics of lung lesions including size, density, and location make the detection of lung nodules on chest X-rays more challenging. However, machine learning methods, including the implementation of deep convolutional neural networks (DCNN), have helped to improve detection.

Facebook's giant "XLM-R" neural network is engineered to work word problems across 100 different languages, including Swahili and Urdu, but it runs up against computing constraints even using 500 of Nvidia's world-class GPUs. With a trend to bigger and bigger machine learning models, state-of-the-art artificial intelligence research continues to run up against the limits of conventional computing technology. Last week they published a report on their invention, XLM-R, a natural language model based on the wildly popular Transformer model from Google. XLM-R is engineered to be able to perform translations between one hundred different languages. It builds upon work that Conneau did earlier this year with Guillaume Lample at Facebook, the creation of the initial XLM.

My previous post on summarising 57 research papers turned out to be quite useful for people working in this field, so it is about time for a sequel. Below you will find short summaries of a number of different research papers published in the areas of Machine Learning and Natural Language Processing in the past couple of years (2017-2019). They cover a wide range of different topics, authors and venues. These are not meant to be reviews showing my subjective opinion, but instead I aim to provide a blunt and concise overview of the core contribution of each publication. Given how many papers are published in our area every year, it is getting more and more difficult to keep track of all of them. The goal of this post is to save some time for both new and experienced readers in the field and allow them to get a quick overview of 74 research papers in about 30 minutes reading time. I set out to post 60 summaries (up from 50 compared to last time). At the end, I also include the summaries for my own published papers since the last iteration (papers 61-74). A transformer architecture that is trained as a language model on a large corpus, then fine-tuned for individual text classification and similarity tasks. Multiple sentences are combined together into a single sequence using delimiters in order to work with the same model.

Their achievement, which will be presented during the "Deep Learning on Supercomputers" workshop at SC19, demonstrates the promise of …

Although AIs are entering new areas every day, a handful of AI laboratories that still focus on artificial intelligence are still consuming large amounts of cash and have made not much progress on AI. According to the documents submitted to the UK Companies Registry in August, only the Alphabet-owned AGI Lab DeepMind lost $570 million in 2018 alone. Another AI Lab, OpenAI, which aims to create AGI, had to abandon its non-profit organisation to find investors in its expensive research. Both labs have achieved extraordinary success, including the creation of robots that can play complex board games and video games. But they are still far from creating artificial intelligence.

Flexible SHAVE Processors: the raw performance of Myriad's SHAVE processors achieve the hundreds of GFLOPS compliments the neural compute engine's hardware fixed-function acceleration. As deep neural network layer types and topologies evolve, the programmability of the SHAVE cores provide the balance between efficiency and future proofing. Massively parallel central memory: deep neural networks create large volumes of intermediate data. Keeping all of this on chip enables our customers to vastly reduce the bandwidth that would otherwise create performance bottlenecks. Myriad X features a proprietary on-chip memory design that minimizes the cost of moving intermediate data – a crucial performance requirement as we see data transfer costs beginning to outstrip data compute costs from an energy perspective.