OpenReview.net; Volodymyr Kuleshov, S. Zayd Enam, Stefano Ermon
from
We propose a neural network-based technique for enhancing the quality of audio signals such as speech or music by transforming inputs encoded at low sampling rates into higher-quality signals with an increased resolution in the time domain. This amounts to generating the missing samples within the low-resolution signal in a process akin to image super-resolution. On standard speech and music datasets, this approach outperforms baselines at 2x, 4x, and 6x upscaling ratios.
The partnership between IBM and one of the world’s top cancer research institutions is falling apart. The project is on hold, MD Anderson confirms, and has been since late last year. MD Anderson is actively requesting bids from other contractors who might replace IBM in future efforts. And a scathing report from auditors at the University of Texas says the project cost MD Anderson more than $62 million and yet did not meet its goals. The report, however, states: “Results stated herein should not be interpreted as an opinion on the scientific basis or functional capabilities of the system in its current state.”
Through my role at Madrona Venture Group advising startups on growth, I meet a lot of marketers using Facebook to acquire and engage with customers. It’s a powerful platform. Unfortunately, many companies are falling short of their goals and are left frustrated with what they see as unrealized potential.
Recently I began a Facebook benchmarking project with the goal of providing marketers in our portfolio with aggregate performance data on different stages of the funnel. This project used 2015 and 2016 data from companies who were targeting consumers. B2B companies were not included in this study, though many have successfully marketed on Facebook and many of the same trends likely apply.
Through this process, I came up with 5 recommendations that companies should consider as they try to get the most out of this large and increasingly competitive channel.
The CS Capacity program was launched in March of 2015 to help address a dramatic increase in undergraduate computer science enrollments that is creating serious resource and pedagogical challenges for many colleges and universities. Over the last two years, a diverse group of universities have been working to develop successful strategies that support the expansion of high-quality CS programs at the undergraduate level. Their work focuses on innovations in teaching and technologies that support scaling while ensuring the engagement of women and underrepresented students. These innovations could provide assistance to many other institutions that are challenged to provide a high-quality educational experience to an increasing number of introductory-level students.
The Data and Software Carpentry staff have been working together to make progress on projects that are important for our community. To help us do this, we’re trying out a new work process based on BaseCamp’s six week work cycle. You can read their blog post if you’re interested in the details of how structuring a work cycle works. We’re picking a small handful of projects to focus on for each six week cycle, with each staff member working on one or two projects. For each project, we’re setting realistic goals we know we can accomplish before the end of the cycle and holding ourselves accountable to meeting those goals. We’re spending the first two weeks of the cycle planning those goals, dividing up the work into teams, and setting timelines to make sure we stay on track.
Data science is a new career for the age of Big Data (whatever that means this week), but you can see that it’s at the intersection of qualities many people have been developing for years. As a graduate of the Science to Data Science (S2DS) summer school, I know people who have come to data science from a wide variety of backgrounds and found a new niche for themselves.
However, I believe there’s something missing from this picture — a vital skill that comes in many forms and needs constant practice and adaption to the situation at hand: communication.
This isn’t just a “soft” or “secondary” skill that’s nice to have. It’s a must-have for good data scientists.
We should be moving our research data into cloud services, or at least backing up our local copies to cloud service.
The cost of storing those hypothetical 28TB of research papers for one year in Google’s cloud platform is $6,800 dollars per year, and if you only need to access a portion of that data, while keeping most of it in a “warm” state, then that cost halves. It’s a bit more than the magical one dollar mark, but it’s possibly within reach of a research budget, and given that most small scale data comes nowhere near this size, the argument for cloud hosting at least a copy of your data becomes quite strong.
A free math camp for middle-school students from New York’s poorest neighborhoods was an effort to increase the number of blacks and Latinos with advanced math degrees.
Despite their tiny size, plankton are a major food source for larger aquatic animals like whales and fish. Astonishingly, plankton also produce half of the planet’s oxygen. Despite their ecological importance, it’s not totally clear how these critters congregate and are pushed around by ocean currents, and how that affects their feeding and mating behaviors.
But that uncertainty is lifting, thanks to a swarm of underwater robots programmed to imitate the movements of plankton. Created by scientists over at the Scripps Institution of Oceanography at the University of California San Diego, they are now using these Miniature Autonomous Underwater Explorers (M-AUE) to better understand how plankton move in three dimensions underwater. The team’s findings could someday also help humanity mitigate threats like oil spills and harmful algal blooms, which can cause massive fish die-offs and make drinking water toxic for humans.
Replication and repeatability are thought by many laypersons to be a shared ideal among many scientists. In practice, few scientific studies are ever replicated. Last year, a survey by Vox.com of 270 scientists found few attempting replication studies because of the difficulty in funding and publishing. Funding agencies pride themselves on sponsoring transformative, breakthrough research – interesting work that, almost by definition, doesn’t repeat (read: replicate) what’s been done before. And journals generally don’t print articles that merely replicate findings that have been previously published; such articles aren’t considered sufficiently interesting.
The results are bad for the practice of science, because the scientific method relies on replication. Without it, it takes a lot longer for erroneous studies to be corrected. But getting things right is not interesting, it’s pedantic.
The first time molecular biologist Greg Hannon flew through a tumour, he was astonished — and inspired. Using a virtual-reality model, Hannon and his colleagues at the University of Cambridge, UK, flew in and out of blood vessels, took stock of infiltrating immune cells and hatched an idea for an unprecedented tumour atlas.
“Holy crap!” he recalls thinking. “This is going to be just amazing.”
On 10 February, the London-based charity Cancer Research UK announced that Hannon’s team of molecular biologists, astronomers and game designers would receive up to £20 million (US$25 million) over the next five years to develop its interactive virtual-reality map of breast cancers.
Carnegie Mellon University’s George Loewenstein presented findings, most in collaboration with students and colleagues in the Social and Decision Sciences Department that challenges traditional economics accounts of how people deal with information. In some cases, motivated by curiosity, people seek out information that has no value for decision making. In other situations, if it threatens to be painful, people avoid information that could inform decisions. And, rather than updating their beliefs rationally, people often defend their beliefs as they would defend material possessions. [video, 1:17:39]
Montreal, Quebec, Canada June 12-16. Come spend a week in Montreal to learn new skills in the computational analysis of literature at McGill University’s .txtLAB. [$$$]
Cambridge, MA We are excited to annouce the 5th annual ComSciCon National Workshop, happening June 8th-10th, 2017 in Cambridge, MA! Applications are now open, and will be accepted through March 1st.
A summer program for security startups
in collaboration with Highland Capital Partners – the program dates are June 12-August 18. We will accept applications starting March 6, continuing on a rolling basis until March 20.
“jq rocks for speedy JSON mangling. Use it to make powerful git clean filters, e.g. when stripping out unwanted cached-data from Jupyter notebooks. You can find the documentation of git ‘clean’ and ‘smudge’ filters buried in the page on git-attributes, or see my example setup.”