Svitkina Z.,Google |
SIAM Journal on Computing | Year: 2011
We introduce several generalizations of classical computer science problems obtained by replacing simpler objective functions with general submodular functions. The new problems include submodular load balancing, which generalizes load balancing or minimum-makespan scheduling, submodular sparsest cut and submodular balanced cut, which generalize their respective graph cut problems, as well as submodular function minimization with a cardinality lower bound. We establish upper and lower bounds for the approximability of these problems with a polynomial number of queries to a function-value oracle. The approximation guarantees that most of our algorithms achieve are of the order of √n/ln n. We show that this is the inherent difficulty of the problems by proving matching lower bounds. We also give an improved lower bound for the problem of approximating a monotone submodular function everywhere. In addition, we present an algorithm for approximating submodular functions with a special structure, whose guarantee is close to the lower bound. Although quite restrictive, the class of functions with this structure includes the ones that are used for lower bounds both by us and in previous work. © 2011 Society for Industrial and Applied Mathematics. Source
The research appears in the journals Marine Policy and Ambio. Marine ecosystems around the world, including small-scale fisheries, are suffering from overfishing. This trend has hurt coral reef ecosystems, where fish species play critical roles in maintaining coral health and abundance. Of particular importance are herbivorous fish that eat macroalgae that overtake reefs if not kept in check. Small-scale fisheries employ 50 million of the world's 51 million fishermen and are responsible for more than half of the annual marine catches around the globe. The majority are in developing countries where limited resources and a local dependence on fishing make effective management difficult. Like many small-scale fishing communities around the world, the village of Buen Hombre in the Dominican Republic is trying to reconcile its fishing with the ecology on which it depends. The community's fishermen mostly capture finfish, with lower harvests of lobster, crabs, octopus and conch. Almost all fishing takes place in coral reef habitats. Fishermen use spearguns, traps, nets, handlines and boats with motors. Spearfishing is done while freediving (using only breath) or with compressors (recycled paint compressors that pressurize air to a diver through a plastic hose). In their first study, the Dartmouth and UCSB researchers found that a mix of factors—notably improvements in fishing gear technology; expanded fishing sites; an increase in fishermen, population, tourism and fish sold to market; an economy increasingly dependent on fishing; and an increased connection between Buen Hombre and the outside world—have contributed to overfishing and the decline of the fishery. In their latest study, the researchers analyzed the fishermen's fishing and social behavior and found that their gear choice largely determined the amount of fish caught (compressor divers have higher catches), how the fishermen divide themselves (such as by membership in the local fishermen's association) and their inability to act collectively to conserve the fishery. "There are both causes for concern and reasons for optimism regarding the future of the Buen Hombre fishery," says co-author Michael Cox, a Dartmouth assistant professor and environmental social scientist who studies community-based natural resource management and environmental governance. "If current trends continue, the fishery is expected to further decline and the reef ecosystem could crash. Shifting the fishery away from this trajectory is not an easy task. Fishermen are very dependent on the fishery and have such little economic leverage within the current market structure that they often have little power to shift their behaviors to more sustainable practices. Despite these challenges, there are several reasons for hope. First, the coral reef ecosystem has not shifted to a completely degraded state. Second, a local nongovernmental organization is establishing incentives for sustainable fishing, helping to organize fishermen, changing the dynamic between fishermen and fish buyers, working to shift the gear types used by fishermen and encouraging government participation."
News Article | January 10, 2016
You know that feeling you get when you’re homesick? That longing feeling in your gut when you miss the people you know and the places you like? If you think that’s bad, imagine being 200 miles above Earth, crammed into a small, apartment-sized room, being forced to see the same handful of people for weeks or even months on end. It’s easy to see why issues would crop up, even for the most well-adjusted. That’s why researchers from Dartmouth and the University of Birmingham are developing virtual reality experiences for astronauts to use while on deep-space missions. The mental health issues resulting from isolation astronauts experience is a big concern, and a virtual escape to the countryside and other scenes of "natural beauty," as the researchers describe them, could offer some relief. “Astronauts are highly functioning and highly-capable adults,” Jay Buckey, a professor of medicine at Dartmouth and former astronaut who flew on space flight STS-90 in 1998, told me. “It’s not like people don’t understand how to get along or anything. Being in an environment like that where you’re in a small, confined space with a very small group of people means that challenges are just going to inevitably arise.” The project, which started last week, is being tested on the closest thing to isolated astronauts you can find on Earth: the residents of Canadian Forces Station Alert (CFS Alert), which is in the northernmost area of inhabited life on the planet, just a few hundred kilometers from the North Pole. "The sun doesn't really come up at CFS Alert during the winter and there aren’t exactly opportunities to do things like going for a walk at the park while you’re there,” Buckey said. Dartmouth College's Digital Arts Leadership and Innovation lab, or DALI, has already developed things like depression and stress management programs in virtual reality with help from NASA. So for this project, the research teams at Dartmouth and Birmingham are focusing on using virtual reality to reconnect astronauts to nature. They hope to find out which virtual reality scenarios work best, and why. “There is a certain plausibility to the idea that we have an inherent connection to nature,” Buckey said. “Just look at the lengths people go through to preserve and experience nature. Natural parks, zoos, hiking, safari excursions. All of those natural experiences are really prized so we want to get to the core of what makes them so popular.” The project is split into two fundamental parts: one focused using virtual reality and 360-degree video, and one focused on recreating picturesque landscapes from Australia and Ireland in virtual reality using 3D graphics. The 360-video project is led by Buckey, while the 3D graphics part of the project (like Virtual Wembury, which you can see in the video above) is led by Robert Stone, director of the human interface technologies team at University of Birmingham. Both teams are working under the same fundamental premise of allowing astronauts to reconnect with nature as a therapeutic exercise. Once crewmembers at CFS Alert view a scene, they're asked to rate it for factors such as realism, immersion, and its calming effects. They'll also be monitored for stress and anxiety levels, sleeping patterns, heart rate, skin conductance, and more. The project is still in its early stages, but crewmembers are expected to try a wide variety of scenes over the course of several weeks. “We use VR reconstructions because we can have full control over lighting, seasonal and weather effects, we can add or remove items as required,” Stone said. “We can also use embedded software to 'track' the end user—where he or she moves, dwells, looks, or what he or she interacts with.” Stone said that with current technology that offers full head-tracking, high definition displays, higher framerates, a wide field of vision, and much more, virtual reality is better than ever, but can it actually trick astronauts into thinking they're somewhere they're not? “With current levels of technology, no,” said Stone. “I have seen many virtual reality head-mounted displays and enclosure-based display systems come and go and not one of them deliver what I would consider to be a genuinely 'immersive' experience, such that the end user actually feels present within an alternative reality. The technology is simply not that capable at the moment.” This assessment doesn’t necessarily mean the technology isn’t worth investing in, just that it isn’t quite capable of literally fooling you into thinking you’re somewhere else. The bright side to that, as Stone explains it, is that also means there is no reason to be concerned about virtual reality addiction or dependence. Buckey and Stone also still don’t know how the known-risks of virtual reality, such as eye-strain and motion sickness, will impact astronauts in zero-gravity. Would it amplify the effects and make disorientation worse, or would the absence of gravity merely negate the effects of virtual reality motion sickness? Hopefully by the time humans start exploring the far reaches of space, virtual reality can figure out the best way to let astronauts take a long walk on the beach by simply putting on a head-mounted display.
News Article | January 26, 2016
Prof. Marvin Minsky, one of the pioneers of artificial intelligence research at the Massachusetts Institute of Technology (MIT), died on Sunday because of cerebral hemorrhage. He was 88 years old. Minsky's death was confirmed by MIT Media Lab founder Nicholas Negroponte through an email he sent to colleagues. "With great great sadness, I have to report that Marvin Minsky died last night," he wrote. "The world has lost one of its greatest minds in science." Negroponte described Minsky as someone who saw the world differently than other people, and that he was able to bring equal measures of deep thinking and humor as a founding member of the MIT Media Lab. Computer scientist Alan Kay, who was a friend and colleague of Minsky's, said that the veteran educator's visions and perspectives helped liberate the computer from its reputation as a "glorified adding machine" and realize its potential as a very powerful tool for human endeavors. Minsky is considered as one of the first scientists to do work on artificial intelligence (AI). In 1956, he attended a Dartmouth symposium, which later became known as the event that helped popularized the concept of AI. In 1959, Minsky and his colleague John McCarthy co-founded MIT's Artificial Intelligence Project, which became known as the Computer Science and Artificial Intelligence Laboratory (CSAIL). He would later write some of the most influential books on the subject such as "Perceptrons" in 1969, "The Society of Mind" in 1986 and "The Emotion Machine" in 2006. Many of Minsky's colleagues view these works as important to understanding what developers have to face in creating intelligence for machines. During his lifetime, Minsky helped develop several technological devices such as the Confocal Scanning Microscope and the SNARC, which is first ever neural network simulator to be created. He also collaborated with fellow educator Edward Fredkin in creating the Triadex Muse synthesizer. Minsky was a member of the U.S. National Academy of Sciences (NAS), U.S. National Academy of Engineering (NAE) and the Extropy Institute's Council of Advisors, among other scientific organizations. Some of the recognitions he received include ACM Turing Award, the IJCAI Research Excellence Award, the Benjamin Franklin Medal, the Robert Wood Prize for Optoelectronics, the Japan Prize and the MIT Killian Award.
Fountains of gas from a handful of remote galaxies all seem to be pointing in roughly the same direction, a new study reports. If the result holds up, it puts a new twist on how galaxies and black holes arise from the larger cosmic web, though some researchers worry that the alignment might just be a chance occurrence. Out of a group of 64 galaxies that are blasting out radio waves, about a dozen are spewing jets of gas that are roughly aligned with one another, astronomers report in the June 11 Monthly Notices of the Royal Astronomical Society Letters. The galactic geysers are powered by supermassive black holes whose magnetic fields launch some infalling debris into intergalactic space. If the geysers are aligned, that means the black holes are all spinning in the same direction. And that means these galaxies, which are spread over roughly a hundred million light-years, might all have been influenced by the larger scaffolding from which they formed. “Naively we expect that shouldn’t happen,” says Ryan Hickox, an astrophysicist at Dartmouth College who was not involved with this study. Black holes, even supermassive ones, are minuscule compared with filaments of galaxies that can span hundreds of millions of light-years. These filaments are the threads along which most matter in the universe congregates, branching through space like a cosmic spider web. Though galaxies live there, they are thought to form and develop independently of what the filaments are doing. A twisting filament should have no influence over what’s happening around one of its resident black holes. And yet that’s the explanation favored by study lead Russ Taylor, an astrophysicist at the University of Cape Town in South Africa. “What we’re seeing is the result of a very large region in the early universe spinning coherently in the same direction,” he says. If that’s true, it adds a “new wrinkle to explain how large-scale structure formed.” Taylor and colleagues found the apparent alignment while probing a patch of sky in the constellation Draco with the Giant Metrewave Radio Telescope in India. They don’t know the distances to the galaxies, but all seem to sit near a galactic supercluster whose light takes about 7.4 billion years to reach Earth — just over half the age of the universe. Other researchers using different techniques have previously reported similar alignments among another set of galaxies (SN: 12/27/14, p. 6). Both studies, though, relied on a small number of galaxies, which means the alignment might not be statistically significant. “If an alignment like this exists, it’s very interesting,” says astrophysicist Michael DiPompeo, also at Dartmouth. “But I’m not super convinced that it’s really there.” While Taylor and colleagues argue that the alignment is not a statistical fluke, DiPompeo did his own calculations that suggest otherwise. He simulated observations of 64 randomly oriented galaxy jets — the computer equivalent of repeatedly dropping a bunch of toothpicks on a table and noting where each was pointed. “I could pretty regularly get patterns that look like this,” he says. It’s also hard to imagine how such an alignment, if it was present as the galaxies formed, could persist for billion of years, he says. “It’s not like [galaxies] form in the early universe and then just sit there blasting these jets.” Galaxies grow by colliding with other galaxies, which can change how the galaxies and their central black holes rotate. Both DiPompeo and Hickox say it’s worth probing other galactic gatherings, though, before dismissing these alignments as a coincidence. If similar orientations appear in many galaxy clusters, then the researchers could be on to something. Hickox would also like to see distances to these galaxies. If it turns out the galaxies sit at wildly varying distances from Earth, he says, then the alignment is less likely to be real. Taylor hopes to do just that. Colleagues are planning observations at other telescopes that will let them determine how far away these galaxies are. And Taylor is gearing up for a more thorough investigation over a much larger patch of sky with a new radio observatory in South Africa called MeerKAT, which should be ready for operation later this year.