PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

Collaborative learning -- for robots

Algorithm lets independent agents collectively produce a machine-learning model without aggregating data

2014-06-25
(Press-News.org) Machine learning, in which computers learn new skills by looking for patterns in training data, is the basis of most recent advances in artificial intelligence, from voice-recognition systems to self-parking cars. It's also the technique that autonomous robots typically use to build models of their environments.

That type of model-building gets complicated, however, in cases in which clusters of robots work as teams. The robots may have gathered information that, collectively, would produce a good model but which, individually, is almost useless. If constraints on power, communication, or computation mean that the robots can't pool their data at one location, how can they collectively build a model?

At the Uncertainty in Artificial Intelligence conference in July, researchers from MIT's Laboratory for Information and Decision Systems will answer that question. They present an algorithm in which distributed agents — such as robots exploring a building — collect data and analyze it independently. Pairs of agents, such as robots passing each other in the hall, then exchange analyses.

In experiments involving several different data sets, the researchers' distributed algorithm actually outperformed a standard algorithm that works on data aggregated at a single location.

"A single computer has a very difficult optimization problem to solve in order to learn a model from a single giant batch of data, and it can get stuck at bad solutions," says Trevor Campbell, a graduate student in aeronautics and astronautics at MIT, who wrote the new paper with his advisor, Jonathan How, the Richard Cockburn Maclaurin Professor of Aeronautics and Astronautics. "If smaller chunks of data are first processed by individual robots and then combined, the final model is less likely to get stuck at a bad solution."

Campbell says that the work was motivated by questions about robot collaboration. But it could also have implications for big data, since it would allow distributed servers to combine the results of their data analyses without aggregating the data at a central location.

"This procedure is completely robust to pretty much any network you can think of," Campbell says. "It's very much a flexible learning algorithm for decentralized networks."

Matching problem

To get a sense of the problem Campbell and How solved, imagine a team of robots exploring an unfamiliar office building. If their learning algorithm is general enough, they won't have any prior notion of what a chair is, or a table, let alone a conference room or an office. But they could determine, for instance, that some rooms contain a small number of chair-shaped objects together with roughly the same number of table-shaped objects, while other rooms contain a large number of chair-shaped objects together with a single table-shaped object.

Over time, each robot will build up its own catalogue of types of rooms and their contents. But inaccuracies are likely to creep in: One robot, for instance, might happen to encounter a conference room in which some traveler has left a suitcase and conclude that suitcases are regular features of conference rooms. Another might enter a kitchen while the coffeemaker is obscured by the open refrigerator door and leave coffeemakers off its inventory of kitchen items.

Ideally, when two robots encountered each other, they would compare their catalogues, reinforcing mutual observations and correcting omissions or overgeneralizations. The problem is that they don't know how to match categories. Neither knows the label "kitchen" or "conference room"; they just have labels like "room 1" and "room 3," each associated with different lists of distinguishing features. But one robot's room 1 could be another robot's room 3.

With Campbell and How's algorithm, the robots try to match categories on the basis of shared list items. This is bound to lead to errors: One robot, for instance, may have inferred that sinks and pedal-operated trashcans are distinguishing features of bathrooms, another that they're distinguishing features of kitchens. But they do their best, combining the lists that they think correspond.

When either of those robots meets another robot, it performs the same procedure, matching lists as best it can. But here's the crucial step: It then pulls out each of the source lists independently and rematches it to the others, repeating this process until no reordering results. It does this again with every new robot it encounters, gradually building more and more accurate models.

Imposing order

This relatively straightforward procedure results from some pretty sophisticated mathematical analysis, which the researchers present in their paper. "The way that computer systems learn these complex models these days is that you postulate a simpler model and then use it to approximate what you would get if you were able to deal with all the crazy nuances and complexities," Campbell says. "What our algorithm does is sort of artificially reintroduce structure, after you've solved that easier problem, and then use that artificial structure to combine the models properly."

In a real application, the robots probably wouldn't just be classifying rooms according to the objects they contain: They'd also be classifying the objects themselves, and probably their uses. But Campbell and How's procedure generalizes to other learning problems just as well.

The example of classifying rooms according to content, moreover, is similar in structure to a classic problem in natural language processing called topic modeling, in which a computer attempts to use the relative frequency of words to classify documents according to topic. It would be wildly impractical to store all the documents on the Web in a single location, so that a traditional machine-learning algorithm could provide a consistent classification scheme for all of them. But Campbell and How's algorithm means that scattered servers could churn away on the documents in their own corners of the Web and still produce a collective topic model.

INFORMATION: END


ELSE PRESS RELEASES FROM THIS DATE:

Are fish near extinction?

2014-06-25
"An end to seafood by 2050?" "Fish to disappear by 2050?" These sensational media headlines were the result of a 2010 report by the United Nations Environment Program, declaring that over-fishing and pollution had nearly emptied the world's fish stocks. That scarcity portends disaster for over a billion people around the world who are dependent on fish for their main source of protein. Now, a new study by Dr. Roi Holzman and Victor China of the Department of Zoology at Tel Aviv University's George S. Wise Faculty of Life Sciences has uncovered the reason why 90% of fish ...

Fifty percent of quality improvement studies fail to change medical practices

2014-06-25
Over the last two decades, nearly half of all initiatives that review and provide feedback to clinicians on healthcare practices show little to no impact on quality of care, according to a new study by Women's College Hospital's Dr. Noah Ivers. The study, published in the Journal of General Internal Medicine, found only 28 per cent of all studies showed an improvement of at least 10 per cent in quality of care over a 25-year period. "Research shows there is a gap between recommended practices and the care patients actually receive," said Dr. Noah Ivers, a family physician ...

Fracking flowback could pollute groundwater with heavy metals

Fracking flowback could pollute groundwater with heavy metals
2014-06-25
VIDEO: This video visualizes the effects of hydrofracking flowback fluid on colloid mobilization in unsaturated sand. Included are the injection of the colloids into the sand column at the beginning of... Click here for more information. ITHACA, N.Y. – The chemical makeup of wastewater generated by "hydrofracking" could cause the release of tiny particles in soils that often strongly bind heavy metals and pollutants, exacerbating the environmental risks during accidental spills, ...

LSTM Researchers demonstrate adaptive potential of hybridization in mosquito species

2014-06-25
Researchers from LSTM have exploited a natural experiment created by insecticidal pressure to determine how the most important malaria vectors - A. gambiae s.s. and A. coluzzii – respond rapidly to environmental change. Working with genome analysis specialists from the Wellcome Trust Sanger Institute and field entomologists in Ghana, LSTM researchers sequenced the genomes of individual wild mosquitoes of each species from southern Ghana. The results, published in Nature Communications, reveal that transfer of a major insecticide resistance mutation (kdr) resulted in replacement ...

World's first magnetic hose created

Worlds first magnetic hose created
2014-06-25
The magnetic hose designed by the researchers consists of a ferromagnetic cylinder covered by a superconductor material, a surprisingly simple design given the complicated theoretical calculations and numerous lab tests it had to undergo. A 14-centimeter prototype was built, which transports the magnetic field from one extreme to the other with a efficiency of 400% in comparison to current methods used to transport these fields. Even with the efficiency of the prototype, researchers theoretically demonstrated that the magnetic hose can be even more efficient if the ferromagnetic ...

New method increases targeted bone volume by 30 percent

2014-06-25
In an important development for the health of elderly people, University of Liverpool researchers have developed a new method to target bone growth. As people age their bones lose density and, especially in women after the menopause, become more brittle. The new method developed by researchers from the University's Institute of Ageing and Chronic Disease offers the possibility of more effective treatment than currently available. Professor Jonathan Jarvis of Liverpool John Moores University designed miniature muscle pacemakers that were used in the University of Liverpool ...

Peer problem solving leads to operational efficiency

2014-06-25
Chestnut Hill, MA (June 25, 2014) - Strength in numbers may not just be a truism for those seeking moral and emotional support, but it also may be an avenue for those seeking customer support. New research shows peer-to-peer problem solving can lessen the need for firms to actually have to contact their supplier for a traditional customer support service call. "This has never been shown before, this notion that people who have full-time day jobs handling support for their companies also take time to answer other people's questions, thereby significantly reducing their ...

Deep brain stimulation improves non motor symptoms in Parkinson's disease

2014-06-25
Amsterdam, NL, 25 June 2014 – Deep brain stimulation (DBS) has become a well-recognized non-pharmacologic treatment that improves motor symptoms of patients with early and advanced Parkinson's disease. Evidence now indicates that DBS can decrease the number and severity of non motor symptoms of patients with Parkinson's disease (PD) as well, according to a review published in the Journal of Parkinson's Disease. "Non motor features are common in PD patients, occur across all disease stages, and while well described, are still under-recognized when considering their huge ...

Reproduction later in life is a marker for longevity in women

2014-06-25
(Boston)--Women who are able to naturally have children later in life tend to live longer and the genetic variants that allow them to do so might also facilitate exceptionally long life spans. A Boston University School of Medicine (BUSM) study published in Menopause: The Journal of the North American Menopause Society, says women who are able to have children after the age of 33 have a greater chance of living longer than women who had their last child before the age of 30. "Of course this does not mean women should wait to have children at older ages in order to ...

The lowdown on triclosan's effects on health and the environment

2014-06-25
Earlier this year, mounting concerns over the potential health effects of triclosan, a common antimicrobial ingredient, prompted Minnesota to ban the germ-killer from consumer soaps statewide starting in 2017. Are these concerns warranted? An article in Chemical & Engineering News (C&EN), the weekly news magazine of the American Chemical Society, investigates. Jyllian Kemsley, a senior editor at C&EN, notes that when it was first patented, triclosan was used as an antimicrobial agent in health care settings. It was a much more benign option as a surgical scrub than the ...

LAST 30 PRESS RELEASES:

Less than half of parents think they have accurate information about bird flu

Common approaches for assessing business impact on biodiversity are powerful, but often insufficient for strategy design

Can a joke make science more trustworthy?

Hiring strategies

Growing consumption of the American eel may lead to it being critically endangered like its European counterpart

KIST develops high-performance sensor based on two-dimensional semiconductor

New study links sleep debt and night shifts to increased infection risk among nurses

Megalodon’s body size and form uncover why certain aquatic vertebrates can achieve gigantism

A longer, sleeker super predator: Megalodon’s true form

Walking, moving more may lower risk of cardiovascular death for women with cancer history

Intracortical neural interfaces: Advancing technologies for freely moving animals

Post-LLM era: New horizons for AI with knowledge, collaboration, and co-evolution

“Sloshing” from celestial collisions solves mystery of how galactic clusters stay hot

Children poisoned by the synthetic opioid, fentanyl, has risen in the U.S. – eight years of national data shows

USC researchers observe mice may have a form of first aid

VUMC to develop AI technology for therapeutic antibody discovery

Unlocking the hidden proteome: The role of coding circular RNA in cancer

Advancing lung cancer treatment: Understanding the differences between LUAD and LUSC

Study reveals widening heart disease disparities in the US

The role of ubiquitination in cancer stem cell regulation

New insights into LSD1: a key regulator in disease pathogenesis

Vanderbilt lung transplant establishes new record

Revolutionizing cancer treatment: targeting EZH2 for a new era of precision medicine

Metasurface technology offers a compact way to generate multiphoton entanglement

Effort seeks to increase cancer-gene testing in primary care

Acoustofluidics-based method facilitates intracellular nanoparticle delivery

Sulfur bacteria team up to break down organic substances in the seabed

Stretching spider silk makes it stronger

Earth's orbital rhythms link timing of giant eruptions and climate change

Ammonia build-up kills liver cells but can be prevented using existing drug

[Press-News.org] Collaborative learning -- for robots
Algorithm lets independent agents collectively produce a machine-learning model without aggregating data