IN THE late 19th century an extraordinary group of women worked at the Harvard College Observatory. Known as “computers”, they charted the position and brightness of stars on a daily basis by applying mathematical formulae to the observations of their male colleagues who watched the sky. Harvard was unique in taking advantage of the burgeoning numbers of educated women in this way. When the observatory’s research was redirected towards photographing the heavens rather than observing them merely by eye, the duties of the “computers” expanded apace. Many of them would go on to extraordinary achievements in astronomy.
The work of Harvard’s female staff was paid for largely by two other women, Anna Palmer Draper and Catherine Wolfe Bruce, heiresses with an enduring interest in astronomy. Dava Sobel, a former science writer for the New York Times who made her name with her bestselling first book, Longitude (1995), has spent several years poring over letters and studying archives in order to tell the story of the women-astronomers and their benefactors.
What’s next? As innovations like artificial intelligence and connected devices become popular, customers are putting cloud components in mobile computing, home games and email marketing campaigns. In other words, the big clouds aim to be everywhere.
All of a sudden, publishers have two voice platforms to worry about.
While publishers big and small – including Digiday! – are still figuring out where Amazon’s voice platform, Alexa, fits into their daily allocation of time and resources, it’s become clear that everybody will have to divide their time.
Google Assistant, the voice platform that animates Google Home, as well as its new phone, the Pixel, has had no trouble making a case that publishers should be building bots for the Assistant too. In just a matter of weeks, it’s attracted publishers ranging from Hearst to the Huffington Post; more than a few of them, including VentureBeat and Genius, decided to take their first steps into voice territory with Google, rather than Amazon.
What are the main trends that will influence the insurance industry in 2017?
We’ve had the exclusive opportunity to find out the view of ten thought leaders on how they see insurance technology in 2017.
With recent progress in graphics, it has become more tractable to train models on synthetic images, potentially avoiding the need for expensive annotations. However, learning from synthetic images may not achieve the desired performance due to a gap between synthetic and real image distributions. To reduce this gap, we propose Simulated+Unsupervised (S+U) learning, where the task is to learn a model to improve the realism of a simulator’s output using unlabeled real data, while preserving the annotation information from the simulator. We develop a method for S+U learning that uses an adversarial network similar to Generative Adversarial Networks (GANs), but with synthetic images as inputs instead of random vectors. We make several key modifications to the standard GAN algorithm to preserve annotations, avoid artifacts and stabilize training: (i) a ‘self-regularization’ term, (ii) a local adversarial loss, and (iii) updating the discriminator using a history of refined images. We show that this enables generation of highly realistic images, which we demonstrate both qualitatively and with a user study. We quantitatively evaluate the generated images by training models for gaze estimation and hand pose estimation. We show a significant improvement over using synthetic images, and achieve state-of-the-art results on the MPIIGaze dataset without any labeled real data.
The development of the embryo into trillions of specialized cells is an intricate genetic dance orchestrated by precisely timed expression of genes. Now a team led by Yale scientists have discovered a way to track the precise bits of RNA that that control this crucial process in a living animal.
This competition is a quest to find new models of global cooperation capable of handling global risks. It will award US$5 million in prizes for the best ideas that re-envision global governance for the 21st century. Deadline for submissions is Wednesday, May 24.
Science encourages us to first attempt to explain new phenomena using our existing knowledge. If this does not work, then new theories must be created. This process is honed in science classes. For example, students rarely receive problems on physics exams that are exactly the same as a homework problem. The expectation is that the students will be able to translate the new exam problems into their existing understanding of physics in order to solve them.
Similarly, in data science, our first instinct is usually to convert a proposed problem into a previously known solution.
This repository contains an implementation of Value Iteration Networks in TensorFlow which won the Best Paper Award at NIPS 2016. This code is based on the original Theano implementation by the authors.