Tag Archives: information

In the researchers' new system, a returning beam of light is mixed with a locally stored beam, and the correlation of their phase, or period of oscillation, helps remove noise caused by interactions with the environment.

Illustration: Jose-Luis Olivares/MIT

Quantum sensor’s advantages survive entanglement breakdown

Preserving the fragile quantum property known as entanglement isn’t necessary to reap benefits.

By Larry Hardesty 


CAMBRIDGE, Mass. – The extraordinary promise of quantum information processing — solving problems that classical computers can’t, perfectly secure communication — depends on a phenomenon called “entanglement,” in which the physical states of different quantum particles become interrelated. But entanglement is very fragile, and the difficulty of preserving it is a major obstacle to developing practical quantum information systems.

In a series of papers since 2008, members of the Optical and Quantum Communications Group at MIT’s Research Laboratory of Electronics have argued that optical systems that use entangled light can outperform classical optical systems — even when the entanglement breaks down.

Two years ago, they showed that systems that begin with entangled light could offer much more efficient means of securing optical communications. And now, in a paper appearing in Physical Review Letters, they demonstrate that entanglement can also improve the performance of optical sensors, even when it doesn’t survive light’s interaction with the environment.

In the researchers' new system, a returning beam of light is mixed with a locally stored beam, and the correlation of their phase, or period of oscillation, helps remove noise caused by interactions with the environment. Illustration: Jose-Luis Olivares/MIT
In the researchers’ new system, a returning beam of light is mixed with a locally stored beam, and the correlation of their phase, or period of oscillation, helps remove noise caused by interactions with the environment.
Illustration Credit: Jose-Luis Olivares/MIT

“That is something that has been missing in the understanding that a lot of people have in this field,” says senior research scientist Franco Wong, one of the paper’s co-authors and, together with Jeffrey Shapiro, the Julius A. Stratton Professor of Electrical Engineering, co-director of the Optical and Quantum Communications Group. “They feel that if unavoidable loss and noise make the light being measured look completely classical, then there’s no benefit to starting out with something quantum. Because how can it help? And what this experiment shows is that yes, it can still help.”

Phased in

Entanglement means that the physical state of one particle constrains the possible states of another. Electrons, for instance, have a property called spin, which describes their magnetic orientation. If two electrons are orbiting an atom’s nucleus at the same distance, they must have opposite spins. This spin entanglement can persist even if the electrons leave the atom’s orbit, but interactions with the environment break it down quickly.

In the MIT researchers’ system, two beams of light are entangled, and one of them is stored locally — racing through an optical fiber — while the other is projected into the environment. When light from the projected beam — the “probe” — is reflected back, it carries information about the objects it has encountered. But this light is also corrupted by the environmental influences that engineers call “noise.” Recombining it with the locally stored beam helps suppress the noise, recovering the information.

The local beam is useful for noise suppression because its phase is correlated with that of the probe. If you think of light as a wave, with regular crests and troughs, two beams are in phase if their crests and troughs coincide. If the crests of one are aligned with the troughs of the other, their phases are anti-correlated.

But light can also be thought of as consisting of particles, or photons. And at the particle level, phase is a murkier concept.

“Classically, you can prepare beams that are completely opposite in phase, but this is only a valid concept on average,” says Zheshen Zhang, a postdoc in the Optical and Quantum Communications Group and first author on the new paper. “On average, they’re opposite in phase, but quantum mechanics does not allow you to precisely measure the phase of each individual photon.”

Improving the odds

Instead, quantum mechanics interprets phase statistically. Given particular measurements of two photons, from two separate beams of light, there’s some probability that the phases of the beams are correlated. The more photons you measure, the greater your certainty that the beams are either correlated or not. With entangled beams, that certainty increases much more rapidly than it does with classical beams.

When a probe beam interacts with the environment, the noise it accumulates also increases the uncertainty of the ensuing phase measurements. But that’s as true of classical beams as it is of entangled beams. Because entangled beams start out with stronger correlations, even when noise causes them to fall back within classical limits, they still fare better than classical beams do under the same circumstances.

“Going out to the target and reflecting and then coming back from the target attenuates the correlation between the probe and the reference beam by the same factor, regardless of whether you started out at the quantum limit or started out at the classical limit,” Shapiro says. “If you started with the quantum case that’s so many times bigger than the classical case, that relative advantage stays the same, even as both beams become classical due to the loss and the noise.”

In experiments that compared optical systems that used entangled light and classical light, the researchers found that the entangled-light systems increased the signal-to-noise ratio — a measure of how much information can be recaptured from the reflected probe — by 20 percent. That accorded very well with their theoretical predictions.

But the theory also predicts that improvements in the quality of the optical equipment used in the experiment could double or perhaps even quadruple the signal-to-noise ratio. Since detection error declines exponentially with the signal-to-noise ratio, that could translate to a million-fold increase in sensitivity.

Source: MIT News Office

The rise and fall of cognitive skills:Neuroscientists find that different parts of the brain work best at different ages.

By Anne Trafton


CAMBRIDGE, Mass–Scientists have long known that our ability to think quickly and recall information, also known as fluid intelligence, peaks around age 20 and then begins a slow decline. However, more recent findings, including a new study from neuroscientists at MIT and Massachusetts General Hospital (MGH), suggest that the real picture is much more complex.

The study, which appears in the XX issue of the journal Psychological Science, finds that different components of fluid intelligence peak at different ages, some as late as age 40.

“At any given age, you’re getting better at some things, you’re getting worse at some other things, and you’re at a plateau at some other things. There’s probably not one age at which you’re peak on most things, much less all of them,” says Joshua Hartshorne, a postdoc in MIT’s Department of Brain and Cognitive Sciences and one of the paper’s authors.

“It paints a different picture of the way we change over the lifespan than psychology and neuroscience have traditionally painted,” adds Laura Germine, a postdoc in psychiatric and neurodevelopmental genetics at MGH and the paper’s other author.

Measuring peaks

Until now, it has been difficult to study how cognitive skills change over time because of the challenge of getting large numbers of people older than college students and younger than 65 to come to a psychology laboratory to participate in experiments. Hartshorne and Germine were able to take a broader look at aging and cognition because they have been running large-scale experiments on the Internet, where people of any age can become research subjects.

Their web sites, gameswithwords.org and testmybrain.org, feature cognitive tests designed to be completed in just a few minutes. Through these sites, the researchers have accumulated data from nearly 3 million people in the past several years.

In 2011, Germine published a study showing that the ability to recognize faces improves until the early 30s before gradually starting to decline. This finding did not fit into the theory that fluid intelligence peaks in late adolescence. Around the same time, Hartshorne found that subjects’ performance on a visual short-term memory task also peaked in the early 30s.

Intrigued by these results, the researchers, then graduate students at Harvard University, decided that they needed to explore a different source of data, in case some aspect of collecting data on the Internet was skewing the results. They dug out sets of data, collected decades ago, on adult performance at different ages on the Weschler Adult Intelligence Scale, which is used to measure IQ, and the Weschler Memory Scale. Together, these tests measure about 30 different subsets of intelligence, such as digit memorization, visual search, and assembling puzzles.

Hartshorne and Germine developed a new way to analyze the data that allowed them to compare the age peaks for each task. “We were mapping when these cognitive abilities were peaking, and we saw there was no single peak for all abilities. The peaks were all over the place,” Hartshorne says. “This was the smoking gun.”

However, the dataset was not as large as the researchers would have liked, so they decided to test several of the same cognitive skills with their larger pools of Internet study participants. For the Internet study, the researchers chose four tasks that peaked at different ages, based on the data from the Weschler tests. They also included a test of the ability to perceive others’ emotional state, which is not measured by the Weschler tests.

The researchers gathered data from nearly 50,000 subjects and found a very clear picture showing that each cognitive skill they were testing peaked at a different age. For example, raw speed in processing information appears to peak around age 18 or 19, then immediately starts to decline. Meanwhile, short-term memory continues to improve until around age 25, when it levels off and then begins to drop around age 35.

For the ability to evaluate other people’s emotional states, the peak occurred much later, in the 40s or 50s.

More work will be needed to reveal why each of these skills peaks at different times, the researchers say. However, previous studies have hinted that genetic changes or changes in brain structure may play a role.

“If you go into the data on gene expression or brain structure at different ages, you see these lifespan patterns that we don’t know what to make of. The brain seems to continue to change in dynamic ways through early adulthood and middle age,” Germine says. “The question is: What does it mean? How does it map onto the way you function in the world, or the way you think, or the way you change as you age?”

Accumulated intelligence

The researchers also included a vocabulary test, which serves as a measure of what is known as crystallized intelligence — the accumulation of facts and knowledge. These results confirmed that crystallized intelligence peaks later in life, as previously believed, but the researchers also found something unexpected: While data from the Weschler IQ tests suggested that vocabulary peaks in the late 40s, the new data showed a later peak, in the late 60s or early 70s.

The researchers believe this may be a result of better education, more people having jobs that require a lot of reading, and more opportunities for intellectual stimulation for older people.

Hartshorne and Germine are now gathering more data from their websites and have added new cognitive tasks designed to evaluate social and emotional intelligence, language skills, and executive function. They are also working on making their data public so that other researchers can access it and perform other types of studies and analyses.

“We took the existing theories that were out there and showed that they’re all wrong. The question now is: What is the right one? To get to that answer, we’re going to need to run a lot more studies and collect a lot more data,” Hartshorne says.

The research was funded by the National Institutes of Health, the National Science Foundation, and a National Defense Science and Engineering Graduate Fellowship.

Source: MIT News Office

Electrical and computer engineering Professor Barry Van Veen wears an electrode net used to monitor brain activity via EEG signals. His research could help untangle what happens in the brain during sleep and dreaming.

Photo Credit: Nick Berard/UW-Madison

Imagination, reality flow in opposite directions in the brain

By Scott Gordon


As real as that daydream may seem, its path through your brain runs opposite reality.

Aiming to discern discrete neural circuits, researchers at the University of Wisconsin–Madison have tracked electrical activity in the brains of people who alternately imagined scenes or watched videos.

“A really important problem in brain research is understanding how different parts of the brain are functionally connected. What areas are interacting? What is the direction of communication?” says Barry Van Veen, a UW-Madison professor of electrical and computer engineering. “We know that the brain does not function as a set of independent areas, but as a network of specialized areas that collaborate.”

Van Veen, along with Giulio Tononi, a UW-Madison psychiatry professor and neuroscientist, Daniela Dentico, a scientist at UW–Madison’s Waisman Center, and collaborators from the University of Liege in Belgium, published results recently in the journalNeuroImage. Their work could lead to the development of new tools to help Tononi untangle what happens in the brain during sleep and dreaming, while Van Veen hopes to apply the study’s new methods to understand how the brain uses networks to encode short-term memory.

During imagination, the researchers found an increase in the flow of information from the parietal lobe of the brain to the occipital lobe — from a higher-order region that combines inputs from several of the senses out to a lower-order region.

Electrical and computer engineering Professor Barry Van Veen wears an electrode net used to monitor brain activity via EEG signals. His research could help untangle what happens in the brain during sleep and dreaming. Photo Credit: Nick Berard/UW-Madison
Electrical and computer engineering Professor Barry Van Veen wears an electrode net used to monitor brain activity via EEG signals. His research could help untangle what happens in the brain during sleep and dreaming.
Photo Credit: Nick Berard/UW-Madison

In contrast, visual information taken in by the eyes tends to flow from the occipital lobe — which makes up much of the brain’s visual cortex — “up” to the parietal lobe.

“There seems to be a lot in our brains and animal brains that is directional, that neural signals move in a particular direction, then stop, and start somewhere else,” says. “I think this is really a new theme that had not been explored.”

The researchers approached the study as an opportunity to test the power of electroencephalography (EEG) — which uses sensors on the scalp to measure underlying electrical activity — to discriminate between different parts of the brain’s network.

Brains are rarely quiet, though, and EEG tends to record plenty of activity not necessarily related to a particular process researchers want to study.

To zero in on a set of target circuits, the researchers asked their subjects to watch short video clips before trying to replay the action from memory in their heads. Others were asked to imagine traveling on a magic bicycle — focusing on the details of shapes, colors and textures — before watching a short video of silent nature scenes.

Using an algorithm Van Veen developed to parse the detailed EEG data, the researchers were able to compile strong evidence of the directional flow of information.

“We were very interested in seeing if our signal-processing methods were sensitive enough to discriminate between these conditions,” says Van Veen, whose work is supported by the National Institute of Biomedical Imaging and Bioengineering. “These types of demonstrations are important for gaining confidence in new tools.”

Source: UW-Madison News

Recommendation theory

Model for evaluating product-recommendation algorithms suggests that trial and error get it right.

By Larry Hardesty

Devavrat Shah’s group at MIT’s Laboratory for Information and Decision Systems (LIDS) specializes in analyzing how social networks process information. In 2012, the group demonstrated algorithms that could predict what topics would trend on Twitter up to five hours in advance; this year, they used the same framework to predict fluctuations in the prices of the online currency known as Bitcoin.

Next month, at the Conference on Neural Information Processing Systems, they’ll present a paper that applies their model to the recommendation engines that are familiar from websites like Amazon and Netflix — with surprising results.

“Our interest was, we have a nice model for understanding data-processing from social data,” says Shah, the Jamieson Associate Professor of Electrical Engineering and Computer Science. “It makes sense in terms of how people make decisions, exhibit preferences, or take actions. So let’s go and exploit it and design a better, simple, basic recommendation algorithm, and it will be something very different. But it turns out that under that model, the standard recommendation algorithm is the right thing to do.”

The standard algorithm is known as “collaborative filtering.” To get a sense of how it works, imagine a movie-streaming service that lets users rate movies they’ve seen. To generate recommendations specific to you, the algorithm would first assign the other users similarity scores based on the degree to which their ratings overlap with yours. Then, to predict your response to a particular movie, it would aggregate the ratings the movie received from other users, weighted according to similarity scores.

To simplify their analysis, Shah and his collaborators — Guy Bresler, a postdoc in LIDS, and George Chen, a graduate student in MIT’s Department of Electrical Engineering and Computer Science (EECS) who is co-advised by Shah and EECS associate professor Polina Golland — assumed that the ratings system had two values, thumbs-up or thumbs-down. The taste of every user could thus be described, with perfect accuracy, by a string of ones and zeroes, where the position in the string corresponds to a particular movie and the number at that location indicates the rating.

Birds of a feather

The MIT researchers’ model assumes that large groups of such strings can be clustered together, and that those clusters can be described probabilistically. Rather than ones and zeroes at each location in the string, a probabilistic cluster model would feature probabilities: an 80 percent chance that the members of the cluster will like movie “A,” a 20 percent chance that they’ll like movie “B,” and so on.

The question is how many such clusters are required to characterize a population. If half the people who like “Die Hard” also like “Shakespeare in Love,” but the other half hate it, then ideally, you’d like to split “Die Hard” fans into two clusters. Otherwise, you’d lose correlations between their preferences that could be predictively useful. On the other hand, the more clusters you have, the more ratings you need to determine which of them a given user belongs to. Reliable prediction from limited data becomes impossible.

In their new paper, the MIT researchers show that so long as the number of clusters required to describe the variation in a population is low, collaborative filtering yields nearly optimal predictions. But in practice, how low is that number?

To answer that question, the researchers examined data on 10 million users of a movie-streaming site and identified 200 who had rated the same 500 movies. They found that, in fact, just five clusters — five probabilistic models — were enough to account for most of the variation in the population.

Missing links

While the researchers’ model corroborates the effectiveness of collaborative filtering, it also suggests ways to improve it. In general, the more information a collaborative-filtering algorithm has about users’ preferences, the more accurate its predictions will be. But not all additional information is created equal. If a user likes “The Godfather,” the information that he also likes “The Godfather: Part II” will probably have less predictive power than the information that he also likes “The Notebook.”

Using their analytic framework, the LIDS researchers show how to select a small number of products that carry a disproportionate amount of information about users’ tastes. If the service provider recommended those products to all its customers, then, based on the resulting ratings, it could much more efficiently sort them into probability clusters, which should improve the quality of its recommendations.

Sujay Sanghavi, an associate professor of electrical and computer engineering at the University of Texas at Austin, considers this the most interesting aspect of the research. “If you do some kind of collaborative filtering, two things are happening,” he says. “I’m getting value from it as a user, but other people are getting value, too. Potentially, there is a trade-off between these things. If there’s a popular movie, you can easily show that I’ll like it, but it won’t improve the recommendations for other people.”

That trade-off, Sanghavi says, “has been looked at in an empirical context, but there’s been nothing that’s principled. To me, what is appealing about this paper is that they have a principled look at this issue, which no other work has done. They’ve found a new kind of problem. They are looking at a new issue.”

Source : MIT News


Carolina’s Laura Mersini-Houghton shows that black holes do not exist

Carolina’s Laura Mersini-Houghton shows that black holes do not exist

 

The term black hole is entrenched in the English language. Can we let it go?

(Chapel Hill, N.C. – Sept. 23, 2014) Black holes have long captured the public imagination and been the subject of popular culture, from Star Trek to Hollywood. They are the ultimate unknown – the blackest and most dense objects in the universe that do not even let light escape. And as if they weren’t bizarre enough to begin with, now add this to the mix: they don’t exist.

By merging two seemingly conflicting theories, Laura Mersini-Houghton, a physics professor at UNC-Chapel Hill in the College of Arts and Sciences, has proven, mathematically, that black holes can never come into being in the first place. The work not only forces scientists to reimagine the fabric of space-time, but also rethink the origins of the universe.

“I’m still not over the shock,” said Mersini-Houghton. “We’ve been studying this problem for a more than 50 years and this solution gives us a lot to think about.”

For decades, black holes were thought to form when a massive star collapses under its own gravity to a single point in space – imagine the Earth being squished into a ball the size of a peanut – called a singularity. So the story went, an invisible membrane known as the event horizon surrounds the singularity and crossing this horizon means that you could never cross back. It’s the point where a black hole’s gravitational pull is so strong that nothing can escape it.

The reason black holes are so bizarre is that it pits two fundamental theories of the universe against each other. Einstein’s theory of gravity predicts the formation of black holes but a fundamental law of quantum theory states that no information from the universe can ever disappear. Efforts to combine these two theories lead to mathematical nonsense, and became known as the information loss paradox.

In 1974, Stephen Hawking used quantum mechanics to show that black holes emit radiation. Since then, scientists have detected fingerprints in the cosmos that are consistent with this radiation, identifying an ever-increasing list of the universe’s black holes.

But now Mersini-Houghton describes an entirely new scenario. She and Hawking both agree that as a star collapses under its own gravity, it produces Hawking radiation. However, in her new work, Mersini-Houghton shows that by giving off this radiation, the star also sheds mass. So much so that as it shrinks it no longer has the density to become a black hole.

Before a black hole can form, the dying star swells one last time and then explodes. A singularity never forms and neither does an event horizon. The take home message of her work is clear: there is no such thing as a black hole.

The paper, which was recently submitted to ArXiv, an online repository of physics papers that is not peer-reviewed, offers exact numerical solutions to this problem and was done in collaboration with Harald Peiffer, an expert on numerical relativity at the University of Toronto. An earlier paper, by Mersini-Houghton, originally submitted to ArXiv in June, was published in the journal Physics Letters B, and offers approximate solutions to the problem.

Experimental evidence may one day provide physical proof as to whether or not black holes exist in the universe. But for now, Mersini-Houghton says the mathematics are conclusive.

Many physicists and astronomers believe that our universe originated from a singularity that began expanding with the Big Bang. However, if singularities do not exist, then physicists have to rethink their ideas of the Big Bang and whether it ever happened.

“Physicists have been trying to merge these two theories – Einstein’s theory of gravity and quantum mechanics – for decades, but this scenario brings these two theories together, into harmony,” said Mersini-Houghton. “And that’s a big deal.”

-Carolina-

Mersini-Houghton’s ArXiv papers:

Approximate solutions:http://arxiv.org/abs/arXiv:1406.1525

Exact solutions:http://arxiv.org/abs/arXiv:1409.1837

Source: UNC News

Visual control of big data

Data-visualization tool identifies sources of aberrant results and recomputes visualizations without them.

By Larry Hardesty


 

CAMBRIDGE, Mass. – In the age of big data, visualization tools are vital. With a single glance at a graphic display, a human being can recognize patterns that a computer might fail to find even after hours of analysis.

But what if there are aberrations in the patterns? Or what if there’s just a suggestion of a visual pattern that’s not distinct enough to justify any strong inferences? Or what if the pattern is clear, but not what was to be expected?

The Database Group at MIT’s Computer Science and Artificial Intelligence Laboratory has released a data-visualization tool that lets users highlight aberrations and possible patterns in the graphical display; the tool then automatically determines which data sources are responsible for which.

It could be, for instance, that just a couple of faulty sensors among dozens are corrupting a very regular pattern of readings, or that a few underperforming agents are dragging down a company’s sales figures, or that a clogged vent in a hospital is dramatically increasing a few patients’ risk of infection.

Big data is big business

Visualizing big data is big business: Tableau Software, which sells a suite of visualization tools, is a $4 billion company. But in creating attractive, informative graphics, most visualization software discards a good deal of useful data.

“If you look at the way people traditionally produce visualizations of any sort, they would have some big, rich data set — that has maybe hundreds of millions of data points, or records — and they would do some reduction of the set to a few hundred or thousands of records at most,” says Samuel Madden, a professor of computer science and engineering and one of the Database Group’s leaders. “The problem with doing that sort of reduction is that you lose information about where those output data points came from relative to the input data set. If one of these data points is crazy — is an outlier, for example — you don’t have any real ability to go back to the data set and ask, ‘Where did this come from and what were its properties?’”

That’s one of the problems solved by the new visualization tool, dubbed DBWipes. For his thesis work, Eugene Wu, a graduate student in electrical engineering and computer science who developed DBWipes with Madden and adjunct professor Michael Stonebraker, designed a novel “provenance tracking” system for large data sets.

If a visualization system summarizes 100 million data entries into 100 points to render on the screen, then each of the 100 points will in some way summarize — perhaps by averaging — 1 million data points. Wu’s provenance-tracking system provides a compact representation of the source of the summarized data so that users can easily trace visualized data back to the source — and conversely, track source data to the pixels that are rendered by it.

The idea of provenance tracking is not new, but Wu’s system is particularly well suited to the task of tracking down outliers in data visualizations. Rather than simply telling the user the million data entries that were used to compute the outliers, it first identifies those that most influenced the outlier values, and summarizes those data entries in human readable terms.

Best paper

Wu and Madden’s work on their “Scorpion” algorithm was selected as one of the best papers of the Very Large Database conference last year. The algorithm tracks down the records responsible for particular aspects of a DBWipes visualization and then efficiently recalculates the visualization to either exclude or emphasize the data they contain.

If some of the points in the visualization suggest a regular pattern, the user can highlight them and mark them as “normal data”; if some of the points disrupt that pattern, the user can highlight them and mark them as “outlier data”; and if the pattern is surprising, the user can draw the anticipated pattern on-screen.

Scorpion then tracks down the provenance of the highlighted points, and filters the provenance down to the subset that most influenced the outliers. Their paper introduces several properties about the specific computation that can be used to develop more efficient algorithms for finding these subsets.

Scorpion, Madden says, was partly motivated by a study conducted by a researcher at a Boston hospital, who noticed that a subset of patients in one of the hospital’s wards was incurring much higher treatment costs than the rest. Any number of factors could have been responsible: the patients’ age and fitness, the severity of their conditions, their particular constellations of symptoms, their health plans, or perhaps something as banal as their proximity to the hospital — nothing could be ruled out.

After six months of work, the researcher concluded that most of the variance in patients’ treatment costs could be explained by a single variable: their doctors. It turned out that three doctors on the hospital staff, in an effort to leave no stone unturned, simply prescribed more interventions than their peers.

As an experiment, Wu and Madden turned Scorpion loose on the researcher’s data. Within five minutes, it had concluded that the data point most strongly correlated with the increase in patients’ treatment costs was the names of their doctors. Because it was combing through a massive data set and, like all big-data search algorithms, had to sacrifice some precision for efficiency, it couldn’t pinpoint just the three doctors identified by the six-month study. But it did produce a list of 10 doctors most likely to be responsible for cost variance, and those three were among them. “You would at least know where to begin looking,” Madden says.

Source:  MIT News Office