Online program wins engineering education award

In collaboration with Boeing and edX, MIT has been honored with the 2017 Excellence in Engineering Education Collaboration Award by the American Society for Engineering Education (ASEE).

The team was chosen for its design and development of a new four-course online professional certification program called Architecture and Systems Engineering: Models and Methods to Manage Complex Systems. The curriculum explores state-of-the-art practices in systems engineering and also demonstrates the value of models in enhancing system engineering functions and augmenting tasks with quantitative analysis.

The program launched last September and ran through March. Nine faculty members from MIT and more than 25 industry experts from Boeing, NASA, IBM, Apple, General Electric, General Motors, and other companies developed content for the courses. More than 1,600 professionals passed the four courses and earned a certificate in the first run of the program. Currently in its second run, the program is now accepting enrollments for September.

“For companies engaged in the development of complex systems, the ability to track the architecture over time is a core competence,” says Bruce Cameron, director of the System Architecture Lab at MIT and faculty director of the program. “As the complexity of the products we produce today increases, engineers face critical challenges managing these systems in the rapidly evolving environment around them. This program prepares the workforce to better face these challenges.”

The program is delivered on the edX platform, with integrated peer-to-peer assessments, group projects, discussion forums, polls, and surveys. In course feedback on the program, more than 93 percent of survey respondents rated the instructors and materials as “good,” “very good,” or “excellent”.

“For my client base, time is the most valuable asset they have. More than money,” explains Michael Fletcher, president of Fletcher Martin Corporation, who earned his professional certificate in March. “When you have a project that's squished into 20 weeks from planning to final completion and there’s a change, a ripple effect happens. Finding ways to minimize that ripple effect and conserve time and money is invaluable. [This program] really built a structured way of thinking that I didn't have before, and brought up a whole new set of ideas. I can't wait to get some models built.”

The development of the program can be traced back to the Space Act Agreement of 2016, when Boeing and NASA joined forces to strengthen engineering and technical leadership capabilities in the United States through innovative educational initiatives. They enlisted MIT and edX to help them create the program. MIT then built a consortium to inform the design of the program, which includes General Electric, Raytheon, Ford, MITRE, and General Motors.

“This partnership with MIT, edX, and NASA blends the expertise of industry, government, and a world-class academic institution to provide a unique educational experience in systems engineering, an area of critical importance to Boeing,” said Greg Hyslop, Boeing's chief technology officer and senior vice president of Engineering, Test and Technology. “That’s a win-win-win for all of us involved, and for the future of aerospace innovation as it’s now applied to learning.”

To earn a certificate, students must complete four courses: Architecture of Complex Systems; Models in Engineering; Model-Based Systems Engineering: Documentation and Analysis; and Quantitative Methods in Systems Engineering. Upon completion, participants are expected to understand and analyze complex systems, perform model management, frame systems architecture as a series of decisions, articulate the benefits and challenges of model-based systems engineering, and demonstrate a comprehensive knowledge of the key aspects of systems engineering.

“The market already offers many educational opportunities around specific tools and new modeling languages. We wanted to offer an overview on why and when to use the tools, in a format that fits into 4-5 hours per week to be compatible with a full-time job,” Cameron says. “The great challenge of system engineering is to foster communication across disciplines — this program builds in a variety of domain examples. ”

Lectures include architectural representations ranging from electrical layout to CAD drawings to functional block diagrams. “That spread is very intentional from our perspective,” Cameron says.

Anant Agarwal, the CEO of edX and an MIT professor, says the success of the program “is a result of edX, MIT, and Boeing’s, combined commitment to providing flexible, highly-engaging digital offerings for professional education at scale and at a fraction of the traditional cost.”

“Together, we are reinventing the way that practicing engineers of hugely complex systems gain access to the new thinking, processes, and tools that help them become more efficient,” Agarwal says.

ASEE, the award sponsor, created the Excellence in Engineering Education Collaboration Awards to demonstrate best practices in collaboration that enhance engineering education. The award competition is open to all ASEE Corporate Member Council organizations for their development of collegiate-level education programs and pre-college programs that generate curiosity and engage students in STEM education.

The award was presented at the 2017 ASEE Annual Conference in Columbus, Ohio, during the Industry Day Plenary Session on June 27.

de MIT News http://ift.tt/2t8XJHa

MIT team races to second in electric car competition

There comes a time in a team's life when it's impossible not to break into a grin. Which is what the MIT Motorsports team did when they snagged a Second Place Overall Spirit of Excellence Award at the Society of Automotive Engineers' Formula SAE Electric competition in Lincoln, Nebraska.

The team earned high finishes in four different categories at the June 24 event, which is part of the SAE Collegiate Design Series. They placed second in the category of Endurance, third in Autocross, third in Skid-Pad, and fifth in Acceleraton. In the Endurance category — which tests acceleration, speed, handling, dynamics, efficiency, and reliability over a 22-kilometer course — the MIT Motorsports team had the fastest lap at 83.7 seconds and overall finished just 17 seconds behind the winners, Penn Electric Racing from the University of Pennsylvania. 

“This has been one of the most difficult things I’ve ever done, and I couldn’t be happier with the results,” team captain Luis Mora said in an email to current and past team members.

MacGyver-style fixes, same-day delivery of key components

The team encountered multiple challenges leading up to the competition.  

Upon arrival in Lincoln, the early-arriving team members discovered that a battery pack contactor had a stripped stud and was unusable. The only replacement contactor was back at the shop at MIT. That left one option: Team member and junior Cameron Ordone drove to the shop, found contactors, tested their resistors, and delivered them to Area 51 CNC shop manager and team advisor Pat McAtamney just minutes before his flight boarded at Logan Airport.

But that wasn't the only problem that required a last-minute fix. First, the vehicle’s front wing did not meet the contest's rules and required modifications. Then a pedal box tab on the frame had a catastrophic mechanical failure during the braking test, so the team had to weld on a sturdier tab and undergo another inspection in order to compete in the day’s events.

“Luis’ leadership of the team was crucial. He made good decisions and stuck by them,” McAtamney said.

Smiling inside the helmet

The team's seniors played a significant role in the group's success. Now graduates, some of them have accepted positions with General Motors and SpaceX and had worked on three consecutive MIT Motorsports electric vehicles. The seniors, Mora said, “rebuilt the team, and now it is our turn to make sure the team progresses forward.”

Senior Will Harvey, who drove the car in the competition, was understandably pleased with the outcome.

“Nothing in the classroom can prepare you for the disappointment of a little water ruining a year of work,” he said, referring to the 2016 competition when the team had issues with the rain test that prevented them from competing in the Endurance category. “But when you turn the final corner and see the checkered flag in a car that you built with 25 friends, you just have to smile a bit inside the helmet.”

When you’re in competition, everything matters, McAtmaney said.

“If we can turn third places into firsts or second places, and do better at Autocross [a one-lap manuvering competition that determines the starting order for the Endurance event], then the team has a good chance of winning next year.”

Junior Cheyenne Hua, the newly-appointed team captain for the upcoming year, said the team's second-place finish this year “proved that we were good.”

“But there is still work to be done before we become exceptional,” she said.

de MIT News http://ift.tt/2u7tgZR

Letter regarding The Engine Working Groups preliminary report

The following email was sent today to the MIT community by Provost Martin A. Schmidt.

To members of the MIT community:

In December 2016, I charged The Engine Working Groups to guide the development of Institute policies and procedures for engaging with The Engine. The Engine, a new external innovation accelerator, was launched by MIT to help start-ups pursuing capital- and time-intensive technologies access patient capital, workspaces, equipment, and services needed to bring solutions from inception to the marketplace. Sixty-two members of the MIT community, including faculty, students, postdocs, and staff, participated in this effort. 

Professor Anantha P. Chandrakasan has led The Engine Working Groups effort and also heads The Engine Advisory Committee, which includes the Working Groups chairs, Vice President and General Counsel Mark DiVincenzo, Executive Director of the Industrial Performance Center Elisabeth Reynolds, and Senior Director for Institute Affairs Glen Comiso.

I now write to share the preliminary report of the Working Groups and to seek your input. Please submit any comments, questions, and suggestions to theenginewg@mit.edu. A final report will be released in early fall.

The Working Groups focused on five areas of MIT engagement with The Engine:

• New Models for Technology Licensing — Chair: Professor Timothy Swager
• Facilities Access — Chair: Professor Martin Culpepper
• Conflict of Interest — Chair: Professor Klavs Jensen
• Visas for MIT Entrepreneurs — Chair: Professor Dick Yue
• MIT’s Innovation Ecosystem — Co-Chairs: Professors Fiona Murray and Vladimir Bulović

This preliminary report is the product of six months of campus-wide engagement, discussions, research, and analysis by the Working Groups. The Advisory Committee submitted the preliminary report to me, Vice President for Research Maria Zuber, and Executive Vice President and Treasurer Israel Ruiz. 

The next step is to collect comments from the community and, after reviewing and incorporating this input, to deliver a final report in early fall and finalize an implementation plan. Finally, while the focus of this report is MIT’s engagement with The Engine, many of the recommendations will enhance overall innovation at the Institute.

I am grateful to Professor Chandrakasan, the chairs of the Working Groups, the Advisory Committee, and the Working Group members for their expertise, time, and service. I am eager to see how these ideas will allow MIT, along with entities such as The Engine, to help young companies develop innovations that positively transform society and help create a better world.

Sincerely,
Martin A. Schmidt

de MIT News http://ift.tt/2t8Pxqn

Bolstering public support for state-level renewable energy policies

Since the 1980s, the United States has often been a world leader in supporting renewable energy technologies at the state and federal level. Thirty-seven states have enacted binding or voluntary renewable portfolio standards (RPS) requiring that a portion of the electricity mix come from renewable sources by a given date. But since 2011, adoption of such standards has slowed, and in the past several years there have been many attempts — some of them successful — to weaken, freeze, or repeal renewable energy laws.

Given the outcome of the 2016 presidential election, increased federal investment in renewable energy is unlikely for the foreseeable future. As a result, state-level renewable energy policies will likely be central to driving new deployment. Past research has shown that public opinion plays a crucial role in facilitating a political consensus around new policies in U.S. states. If that’s true for renewable energy policies, then people’s views may have a major influence on future actions taken by their states. 

For the past three years, MIT Associate Professor Christopher Warshaw of the Department of Political Science and Leah Stokes SM ’15, PhD ’15, now an assistant professor of political science at the University of California at Santa Barbara, have been examining the interaction between public opinion and renewable energy policymaking. First, is there evidence that public opinion and energy policy align within a particular state? And second, what determines that public opinion? For example, can the design of a given RPS policy or how it’s presented to the public — that is, how it’s portrayed or framed — increase or decrease support for the policy?

Now, an analysis by Warshaw and Stokes finds that state legislators are, in fact, broadly responsive to public opinion in this policy arena. And based on data from a public opinion survey, the researchers offer practical advice on how to bolster public support for renewable policies. Their findings are published today in the journal Nature Energy.

Public opinion and renewable energy policy, state by state 

To begin investigating their questions, Warshaw and Stokes turned to data gathered by the Cooperative Congressional Election Study, a major survey supported by 56 universities, including MIT, that has its origins in a survey first funded by the MIT Energy Initiative a decade ago. In the 2014 cooperative survey, 56,200 people were asked whether they supported an RPS policy that “requires the use of a minimum amount of renewable fuels (wind, solar, and hydroelectric) in the generation of electricity, even if electricity prices increase a little.”

Using the 2014 survey data, Warshaw and Stokes explored the relationship between public opinion and policy on a state-by-state basis. Their analysis showed that in most states a majority of the public supports renewable energy requirements — although frequently by a narrow margin. In addition, public support within each state is strongly correlated with the RPS policy now in effect. Thirty-seven states plus the District of Columbia have RPS policies that are congruent with the views of a majority of their citizens, leaving only 13 that don’t. All 13 states where more than 60 percent of the public supports an RPS have a binding RPS policy, with varying levels of ambitiousness. As public support drops close to or below 50 percent, states are much less likely to have a binding RPS.

“Overall, these findings suggest that state legislators are broadly responsive to public opinion on this issue,” says Warshaw. “If public support for renewable energy policies increased, we could expect to see more renewable energy laws.” 

A new experiment 

In other areas of policymaking, research has shown that exactly how a policy is designed and presented can significantly impact whether the public supports or opposes it. Thus, it’s possible that certain details of RPS policies could be swaying public opinion. “We needed to gauge how the design and framing of renewable energy policies may affect people’s support for them across the states,” says Warshaw. He and Stokes set out to design a survey experiment that would give them insight into what drives people’s opinions of renewable energy policies. 

They knew many factors could influence support for an RPS policy — from possible changes in electric bills to impacts on employment opportunities. A simple survey experiment might involve randomizing one such attribute at a time. For example, one group could be told that the new policy will increase residential electric bills, and the group’s response could then be compared to that of a control group that receives no information about added costs. 

But the attributes of interest here are independent — they have no impact on one another — so the researchers could investigate all of them simultaneously. With this approach, the effects of the different attributes are all measured on the same scale. When the results are in, it’s easy to see which factors are most important and warrant special attention or concern. 

In the new survey, all recipients received a central statement posing the possibility of the recipient’s state adopting a new RPS bill requiring that the state meet 35 percent of its electricity needs with renewable energy sources by the year 2025. Along with that description, they received a variety of additional statements about specific attributes of the bill, randomly distributed among the survey recipients. For each attribute, some (randomly selected) people received no added information, thereby serving as the control group in the experiment. 

Warshaw and Stokes received replies from about 2,500 respondents. They then performed a statistical analysis on all the data to determine how much information on each of the attributes changed people’s views of the basic RPS policy from those of the control group. 

Economic incentives — costs and jobs 

The results show that an increase in residential energy costs has a far greater impact on the outcome than any of the other attributes. Adding $2 to an electricity bill decreased support for an RPS policy by about 6 percent, while a $10 increase decreased support by fully 13 percent. Those changes are large enough to flip majority public opinion within some states from supporting to opposing RPS policies. In the $2 case, 13 states shifted from supporting to opposing; in the $10 case, 33 states moved to the opposing side. 

The possible impact on jobs is another big factor — one that can push support either way. Being told that the bill won’t create any jobs prompted 3.2 percent of respondents to oppose the bill. With that change, five states flipped from majority support to majority opposition. On the other hand, learning that the RPS policy will probably create several thousand jobs caused 7 percent of respondents to support the bill, a change that flipped eight states from majority opposition to majority support. “So if people think these policies will create a lot of jobs, public support increases enough to lead almost every state — except possibly the most conservative ones — to support RPS policies,” Stokes notes. 

The results provide some interesting clues about what people believe now. For example, the response to added costs suggests that many people think renewables won’t — or shouldn’t — cost them anything extra. The prospect of a $2 increase in their electricity bill prompts a shift toward opposition. If people started out thinking renewable standards would cost them something, adding just $2 to the bill probably wouldn’t have elicited such a change. 

The negative response to learning that the new policy will bring no extra jobs conveys a different message. “It may suggest that in the absence of any added information, people think the new bill will lead to a small increase in jobs — which frankly is generally about right,” says Warshaw. Once again, the experiment uncovered starting assumptions that people may have — perhaps without knowing it.

Environmental impacts 

Another reason to support using renewable energy may be the promise of environmental benefits. The survey tested that idea by telling some respondents that increasing renewable energy will reduce harmful air pollution in their state, including toxins such as mercury. Learning that air pollution will go down brings almost as large a response as learning that employment will go up: 6.7 percent of people move to the supporting side. “So emphasizing either job creation or air quality benefits could cause eight of the 10 states where a majority now opposes RPS bills — and where RPS policies largely do not exist — to flip to a majority in support,” says Stokes. 

Interestingly, linking RPS policies to climate change had no impact on public support. The survey included various statements about the effects of RPS policies on greenhouse gas emissions and about whether or not supporters and opponents believe climate change to be a serious problem. While the added information increased support slightly, the change wasn’t large enough to be statistically significant. 

Warshaw believes that the lack of impact isn’t because people don’t know or care about climate change. “I think it’s because they already have a pretty strong view on the connection between renewable energy policies and climate change,” he says. “Their view is already baked in, so you can’t frame the question in a way that triggers a change.” 

Partisan support 

One more factor of interest is the role played by elites in U.S. political parties. Some research suggests that partisanship isn’t important for energy policy, even though it has been shown to influence public support in other policy domains. So the researchers added some partisan cues. 

They found that when people were told that Democratic legislators support the RPS policy, public support increased by 2.4 percent, and three states flipped from majority opposition to majority support. When respondents were told that Republican legislators support it, public support increased by 5.5 percent, and seven states flip to majority support. Interestingly, the results show that if an elite affiliated with one political party supports the RPS policy, there is no statistically significant decrease in support by respondents affiliated with the other party. 

Warshaw believes that support by partisan elites can have a big impact in part because people’s views on renewable energy “aren’t super-strongly formed,” he says. “On policies they don’t know much about, people look to their elected officials to tell them what the right thing to believe is. There’s considerable political science evidence that that’s true.” 

Stokes notes that while none of the statements relating to climate change seemed to influence public opinion in the survey, in the absence of a coherent federal policy, state-level RPS policies may actually prove the most effective means of securing climate benefits. That prospect underscores the need for continuing public engagement during the decades-long process of weaning the U.S. energy system off fossil fuels. 

This research was supported by the MIT Energy Initiative Seed Fund Program. While at MIT, Leah Stokes was a 2010-2011 Siemens-MIT Energy Fellow and a 2013-2014 Martin Family Sustainability Fellow. Logistical support was provided by the MIT Political Experiments Research Lab.

de MIT News http://ift.tt/2suvfFJ

J-PAL North America partners with local governments to tackle homelessness and reduce incarceration

J-PAL North America, a research center at MIT, announced today that it will partner with three city and county governments to evaluate promising solutions to homelessness and other important policy challenges facing state and local governments in the U.S. The City of Baltimore, Maryland; King County, Washington; and Santa Clara County, California, will work with J-PAL North America staff and leading academic researchers to test programs designed to help vulnerable individuals find and keep housing and to reduce jail time and recidivism for low-level offenders.

“We’re excited to partner with Baltimore, King County, and Santa Clara County as they take on some of the most important policy issues facing state and local leaders,” says Melissa Kearney, professor of economics at the University of Maryland, non-resident senior fellow at the Brookings Institution, and co-chair of the J-PAL State and Local Innovation Initiative. “Through rigorous research, we can help governments identify the most effective and cost-effective solutions to these challenges.”

“The goal of this initiative is to give government leaders tools to better understand the impact of their policies and programs,” adds Jonathan Guryan, initiative co-chair and associate professor of human development and social policy and economics in the School of Education and Social Policy and Faculty Fellow in the Institute for Policy Research at Northwestern University. “We hope that these partnerships will help these governments make better informed decisions and ultimately improve the lives of the people they serve.”

Baltimore’s Mayor’s Office of Human Services is partnering with J-PAL North America to test innovative approaches to reducing homelessness among unaccompanied youth. Baltimore has seen a 21 percent rise in the number of homeless young people aged 18 to 24 in the last three years, and the city’s few programs designed for unaccompanied youth are at full capacity. As the city works to increase investment to prevent youth homelessness, it’s critical that new resources are directed toward the most effective interventions. To that end, the Mayor’s Office of Human Services will partner with J-PAL North America to develop an evaluation of housing and supportive services for homeless youth, with the goal of reducing the length of time youth are homeless and helping them achieve long-term stability and independence.

“The City of Baltimore is thrilled to be selected as a partner city for the J-PAL State and Local Innovation Initiative,” said Baltimore Mayor Catherine Pugh. “This opportunity to work with top experts and researchers will put Baltimore at the forefront of national efforts to develop evidence-based services and interventions for unaccompanied youth experiencing homelessness. Over the past two years, we’ve worked to better serve homeless youth by conducting counts and surveys, investing in new housing opportunities and streamlining our systems of care. We look forward to working with J-PAL to evaluate what approaches are working best and build on our success.”

King County’s Department of Community and Human Services will partner with J-PAL North America to develop evaluations of two important programs — one aimed at preventing and reducing homelessness and another geared towards decreasing jail time and reducing recidivism for low-level offenders. The county’s Department of Community and Human Services coordinates services with community-based agencies to reach the region’s most vulnerable populations, such as people experiencing homelessness, mental illness, and substance use disorders. King County will partner with J-PAL North America and the Wilson Sheehan Lab for Economic Opportunities (LEO) to develop an evaluation of the Youth and Family Homelessness Prevention Initiative, which provides flexible funding and case management to families and unaccompanied youth. King County will also partner with J-PAL North America and LEO to evaluate the effectiveness of the Law Enforcement Assisted Diversion program, a pre-booking diversion program that redirects low-level offenders engaged in drug or prostitution activity to case management and community services.

“King County is continually improving our evidence-based policymaking to achieve better outcomes for our residents,” said King County Executive Dow Constantine. “I am thrilled King County was selected for J-PAL's State and Local Innovation Initiative. This partnership will allow us to better evaluate our homelessness prevention efforts and reduce involvement in our criminal justice system. We look forward to working with Notre Dame's Lab for Economic Opportunities and learning from other jurisdictions and J-PAL's network.”

Santa Clara County’s Office of Supportive Housing will partner with J-PAL North America to develop an evaluation of potential solutions to address homelessness among single adults. Despite its affluence, Santa Clara County has one of the largest homeless populations in the country, numbering 6,556 individuals in 2015 — 70 percent of whom have been found to be living in unsheltered conditions such as on the streets, under bridges, in cars, or in abandoned buildings. The county’s Office of Supportive Housing leads efforts to supply affordable housing for extremely low-income and special-needs households. Through a partnership with J-PAL North America and LEO, Santa Clara aims to build needed evidence of the impact of its housing interventions, not only on homelessness but also on public health and criminal justice outcomes, with the ultimate goal of ending homelessness within five years.

“Our partnership with the University of Notre Dame and MIT demonstrates our ongoing commitment to implementing the most cost-effective and evidence-based solutions to reducing homelessness in Santa Clara County,” says Ky Lee, director of Santa Clara’s Office of Supportive Housing.

All of these programs are expected to attract more applicants than they can serve. This creates an opportunity to assign slots to the program randomly, which is seen as a fair way to allocate scarce resources and allows researchers to rigorously evaluate the program.

“We were impressed by the strong commitment of the leaders in these governments to using data and rigorous evaluations to inform their policy decisions,” says Mary Ann Bates, deputy executive director of J-PAL North America and initiative co-chair. “Together with the other applicants to this initiative, these leaders represent an important movement toward evidence-based policymaking in the United States.”

The three jurisdictions will join five governments selected last year in the inaugural round of the J-PAL State and Local Innovation Initiative: Pennsylvania, Philadelphia, Puerto Rico, Rochester, and South Carolina. This initiative supports state and local governments in generating new and widely applicable lessons about which social programs work, which work best, and why. These government partners and J-PAL North America are working to bridge the gap between academia and policy, and translate research into action that has a real impact on people’s lives.

For more information, contact initiative manager Julia Chabrier. To learn more, or to sign up to receive updates on the J-PAL State and Local Innovation Initiative, visit http://ift.tt/1I2hzTW.

de MIT News http://ift.tt/2trM8oZ

Peering into neural networks

Neural networks, which learn to perform computational tasks by analyzing large sets of training data, are responsible for today’s best-performing artificial intelligence systems, from speech recognition systems, to automatic translators, to self-driving cars.

But neural nets are black boxes. Once they’ve been trained, even their designers rarely have any idea what they’re doing — what data elements they’re processing and how.

Two years ago, a team of computer-vision researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) described a method for peering into the black box of a neural net trained to identify visual scenes. The method provided some interesting insights, but it required data to be sent to human reviewers recruited through Amazon’s Mechanical Turk crowdsourcing service.

At this year’s Computer Vision and Pattern Recognition conference, CSAIL researchers will present a fully automated version of the same system. Where the previous paper reported the analysis of one type of neural network trained to perform one task, the new paper reports the analysis of four types of neural networks trained to perform more than 20 tasks, including recognizing scenes and objects, colorizing grey images, and solving puzzles. Some of the new networks are so large that analyzing any one of them would have been cost-prohibitive under the old method.

The researchers also conducted several sets of experiments on their networks that not only shed light on the nature of several computer-vision and computational-photography algorithms, but could also provide some evidence about the organization of the human brain.

Neural networks are so called because they loosely resemble the human nervous system, with large numbers of fairly simple but densely connected information-processing “nodes.” Like neurons, a neural net’s nodes receive information signals from their neighbors and then either “fire” — emitting their own signals — or don’t. And as with neurons, the strength of a node’s firing response can vary.

In both the new paper and the earlier one, the MIT researchers doctored neural networks trained to perform computer vision tasks so that they disclosed the strength with which individual nodes fired in response to different input images. Then they selected the 10 input images that provoked the strongest response from each node.

In the earlier paper, the researchers sent the images to workers recruited through Mechanical Turk, who were asked to identify what the images had in common. In the new paper, they use a computer system instead.

“We catalogued 1,100 visual concepts — things like the color green, or a swirly texture, or wood material, or a human face, or a bicycle wheel, or a snowy mountaintop,” says David Bau, an MIT graduate student in electrical engineering and computer science and one of the paper’s two first authors. “We drew on several data sets that other people had developed, and merged them into a broadly and densely labeled data set of visual concepts. It’s got many, many labels, and for each label we know which pixels in which image correspond to that label.”

The paper’s other authors are Bolei Zhou, co-first author and fellow graduate student; Antonio Torralba, MIT professor of electrical engineering and computer science; Aude Oliva, CSAIL principal research scientist; and Aditya Khosla, who earned his PhD as a member of Torralba’s group and is now the chief technology officer of the medical-computing company PathAI.

The researchers also knew which pixels of which images corresponded to a given network node’s strongest responses. Today’s neural nets are organized into layers. Data are fed into the lowest layer, which processes them and passes them to the next layer, and so on. With visual data, the input images are broken into small chunks, and each chunk is fed to a separate input node.

For every strong response from a high-level node in one of their networks, the researchers could trace back the firing patterns that led to it, and thus identify the specific image pixels it was responding to. Because their system could frequently identify labels that corresponded to the precise pixel clusters that provoked a strong response from a given node, it could characterize the node’s behavior with great specificity.

The researchers organized the visual concepts in their database into a hierarchy. Each level of the hierarchy incorporates concepts from the level below, beginning with colors and working upward through textures, materials, parts, objects, and scenes. Typically, lower layers of a neural network would fire in response to simpler visual properties — such as colors and textures — and higher layers would fire in response to more complex properties.

But the hierarchy also allowed the researchers to quantify the emphasis that networks trained to perform different tasks placed on different visual properties. For instance, a network trained to colorize black-and-white images devoted a large majority of its nodes to recognizing textures. Another network, when trained to track objects across several frames of video, devoted a higher percentage of its nodes to scene recognition than it did when trained to recognize scenes; in that case, many of its nodes were in fact dedicated to object detection.

One of the researchers’ experiments could conceivably shed light on a vexed question in neuroscience. Research involving human subjects with electrodes implanted in their brains to control severe neurological disorders has seemed to suggest that individual neurons in the brain fire in response to specific visual stimuli. This hypothesis, originally called the grandmother-neuron hypothesis, is more familiar to a recent generation of neuroscientists as the Jennifer-Aniston-neuron hypothesis, after the discovery that several neurological patients had neurons that appeared to respond only to depictions of particular Hollywood celebrities.

Many neuroscientists dispute this interpretation. They argue that shifting constellations of neurons, rather than individual neurons, anchor sensory discriminations in the brain. Thus, the so-called Jennifer Aniston neuron is merely one of many neurons that collectively fire in response to images of Jennifer Aniston. And it’s probably part of many other constellations that fire in response to stimuli that haven’t been tested yet.

Because their new analytic technique is fully automated, the MIT researchers were able to test whether something similar takes place in a neural network trained to recognize visual scenes. In addition to identifying individual network nodes that were tuned to particular visual concepts, they also considered randomly selected combinations of nodes. Combinations of nodes, however, picked out far fewer visual concepts than individual nodes did — roughly 80 percent fewer.

“To my eye, this is suggesting that neural networks are actually trying to approximate getting a grandmother neuron,” Bau says. “They’re not trying to just smear the idea of grandmother all over the place. They’re trying to assign it to a neuron. It’s this interesting hint of this structure that most people don’t believe is that simple.”

de MIT News http://ift.tt/2ts5bz7

West Garage scheduled to close September 2017

On Sept. 15, 2017, the West Garage (W45), located on Vassar Street across from the Johnson Athletic Center and Steinbrenner Stadium, is scheduled to close in preparation for the construction of a new undergraduate residence hall on the site. The garage is expected to remain available on a limited basis through mid-November for special events and athletic events. In late fall and early winter, the construction project team will conduct exploratory work and will prepare the structure for demolition, which is slated to take place in early 2018.

To accommodate MIT permit holders who currently use the roughly 370 parking spaces in the West Garage, the Parking and Transportation Office has developed a solution based on assigning permit holders to other parking areas on campus and incorporating a new Attendant Assist parking program at Stata Garage.

Permit renewal process opens June 30

On June 30, the parking permit renewal process will open for the 2017-2018 year. Permit holders who were assigned to the West Garage for the previous year will now be assigned to a new parking area. These parking areas will include Northwest, Northeast (Stata), Riverside, and North. When individuals log into the parking site through Atlas to renew their permits, the new assignments will already be entered into the system.

“The Parking and Transportation team has reviewed parking assignments carefully,” says Tom Giannino, operations manager for the Parking and Transportation Office. “For each individual, we assessed the proximity from parking area to office with the goal of minimizing walking distances as much as possible.”

Currently, West Garage is used frequently for athletic events and special events parking, and the garage will remain available for these uses through mid-November. Starting at the end of November, the Albany Garage will become the primary parking facility for athletic events and special events, supplemented by Northwest area lots.

Launching the Attendant Assist program at Stata Garage

With the goal of keeping parkers closer to their offices, the team is planning to increase the availability of parking in Stata Garage by introducing the Attendant Assist program. This program is being implemented at Stata to allow more parkers access to a centrally located garage.

Once the regular parking spaces in Stata are all full, parkers — with assistance from a customer service representative (CSR) and parking attendants — will be directed to park in aisle spaces and will leave their keys with the attendants. Keys will be stored in a locked box at each aisle, and attendants will move cars as needed to assist parkers moving in and out of spaces. The process will be overseen by an on-site supervisor.

During the recent renovation of Building E70 at 1 Broadway, the Attendant Assist program was employed successfully to manage parking spaces in the E70 garage and keep the garage available to parkers during construction. The program at Stata is expected to benefit from the experience of the team managing this solution.

Evolving commuter options for an evolving campus

As part of the Kendall Square project, a new underground parking garage is scheduled to open in 2020, adding 500 parking spaces back into the inventory. In the meantime, as MIT moves forward with enhancements to the campus to meet research and residential needs, the Parking and Transportation Office and Office of Campus Planning will continue to seek more options for members of the community as they navigate to and from the campus.

“Our goal is to ease the burden on our parkers as much as possible,” says John DiFava, director of campus services and chief of police at MIT. “We’re implementing a solution that has met with success elsewhere, and we will adapt the program once we observe how it is working for our community.”

This past year, the Parking and Transportation Office, in partnership with the MIT Office of Sustainability and the MIT Transit Lab, launched the Access MIT program, a suite of enhanced commuter benefits designed to encourage sustainable transportation practices. Access MIT makes it easier for MIT community members to seek lower-carbon transportation options, such as commuting by bike or via public transportation instead of by car, supporting MIT’s goal of reducing parking demand on campus 10 percent by 2018. The program provides eligible employees with benefits such as free MBTA subway and local bus passes, commuter rail subsidies, and subsidies for parking at MBTA stations. To date, Access MIT has succeeded in reducing parking demand on campus by almost 5 percent.

MIT individuals with questions about the parking program may contact the Parking and Transportation Office at mitparking@mit.edu or 617-258-6510.

de MIT News http://ift.tt/2slyDr9