More stories

  • in

    Dealing with the limitations of our noisy world

    Tamara Broderick first set foot on MIT’s campus when she was a high school student, as a participant in the inaugural Women’s Technology Program. The monthlong summer academic experience gives young women a hands-on introduction to engineering and computer science.

    What is the probability that she would return to MIT years later, this time as a faculty member?

    That’s a question Broderick could probably answer quantitatively using Bayesian inference, a statistical approach to probability that tries to quantify uncertainty by continuously updating one’s assumptions as new data are obtained.

    In her lab at MIT, the newly tenured associate professor in the Department of Electrical Engineering and Computer Science (EECS) uses Bayesian inference to quantify uncertainty and measure the robustness of data analysis techniques.

    “I’ve always been really interested in understanding not just ‘What do we know from data analysis,’ but ‘How well do we know it?’” says Broderick, who is also a member of the Laboratory for Information and Decision Systems and the Institute for Data, Systems, and Society. “The reality is that we live in a noisy world, and we can’t always get exactly the data that we want. How do we learn from data but at the same time recognize that there are limitations and deal appropriately with them?”

    Broadly, her focus is on helping people understand the confines of the statistical tools available to them and, sometimes, working with them to craft better tools for a particular situation.

    For instance, her group recently collaborated with oceanographers to develop a machine-learning model that can make more accurate predictions about ocean currents. In another project, she and others worked with degenerative disease specialists on a tool that helps severely motor-impaired individuals utilize a computer’s graphical user interface by manipulating a single switch.

    A common thread woven through her work is an emphasis on collaboration.

    “Working in data analysis, you get to hang out in everybody’s backyard, so to speak. You really can’t get bored because you can always be learning about some other field and thinking about how we can apply machine learning there,” she says.

    Hanging out in many academic “backyards” is especially appealing to Broderick, who struggled even from a young age to narrow down her interests.

    A math mindset

    Growing up in a suburb of Cleveland, Ohio, Broderick had an interest in math for as long as she can remember. She recalls being fascinated by the idea of what would happen if you kept adding a number to itself, starting with 1+1=2 and then 2+2=4.

    “I was maybe 5 years old, so I didn’t know what ‘powers of two’ were or anything like that. I was just really into math,” she says.

    Her father recognized her interest in the subject and enrolled her in a Johns Hopkins program called the Center for Talented Youth, which gave Broderick the opportunity to take three-week summer classes on a range of subjects, from astronomy to number theory to computer science.

    Later, in high school, she conducted astrophysics research with a postdoc at Case Western University. In the summer of 2002, she spent four weeks at MIT as a member of the first class of the Women’s Technology Program.

    She especially enjoyed the freedom offered by the program, and its focus on using intuition and ingenuity to achieve high-level goals. For instance, the cohort was tasked with building a device with LEGOs that they could use to biopsy a grape suspended in Jell-O.

    The program showed her how much creativity is involved in engineering and computer science, and piqued her interest in pursuing an academic career.

    “But when I got into college at Princeton, I could not decide — math, physics, computer science — they all seemed super-cool. I wanted to do all of it,” she says.

    She settled on pursuing an undergraduate math degree but took all the physics and computer science courses she could cram into her schedule.

    Digging into data analysis

    After receiving a Marshall Scholarship, Broderick spent two years at Cambridge University in the United Kingdom, earning a master of advanced study in mathematics and a master of philosophy in physics.

    In the UK, she took a number of statistics and data analysis classes, including her first class on Bayesian data analysis in the field of machine learning.

    It was a transformative experience, she recalls.

    “During my time in the U.K., I realized that I really like solving real-world problems that matter to people, and Bayesian inference was being used in some of the most important problems out there,” she says.

    Back in the U.S., Broderick headed to the University of California at Berkeley, where she joined the lab of Professor Michael I. Jordan as a grad student. She earned a PhD in statistics with a focus on Bayesian data analysis. 

    She decided to pursue a career in academia and was drawn to MIT by the collaborative nature of the EECS department and by how passionate and friendly her would-be colleagues were.

    Her first impressions panned out, and Broderick says she has found a community at MIT that helps her be creative and explore hard, impactful problems with wide-ranging applications.

    “I’ve been lucky to work with a really amazing set of students and postdocs in my lab — brilliant and hard-working people whose hearts are in the right place,” she says.

    One of her team’s recent projects involves a collaboration with an economist who studies the use of microcredit, or the lending of small amounts of money at very low interest rates, in impoverished areas.

    The goal of microcredit programs is to raise people out of poverty. Economists run randomized control trials of villages in a region that receive or don’t receive microcredit. They want to generalize the study results, predicting the expected outcome if one applies microcredit to other villages outside of their study.

    But Broderick and her collaborators have found that results of some microcredit studies can be very brittle. Removing one or a few data points from the dataset can completely change the results. One issue is that researchers often use empirical averages, where a few very high or low data points can skew the results.

    Using machine learning, she and her collaborators developed a method that can determine how many data points must be dropped to change the substantive conclusion of the study. With their tool, a scientist can see how brittle the results are.

    “Sometimes dropping a very small fraction of data can change the major results of a data analysis, and then we might worry how far those conclusions generalize to new scenarios. Are there ways we can flag that for people? That is what we are getting at with this work,” she explains.

    At the same time, she is continuing to collaborate with researchers in a range of fields, such as genetics, to understand the pros and cons of different machine-learning techniques and other data analysis tools.

    Happy trails

    Exploration is what drives Broderick as a researcher, and it also fuels one of her passions outside the lab. She and her husband enjoy collecting patches they earn by hiking all the trails in a park or trail system.

    “I think my hobby really combines my interests of being outdoors and spreadsheets,” she says. “With these hiking patches, you have to explore everything and then you see areas you wouldn’t normally see. It is adventurous, in that way.”

    They’ve discovered some amazing hikes they would never have known about, but also embarked on more than a few “total disaster hikes,” she says. But each hike, whether a hidden gem or an overgrown mess, offers its own rewards.

    And just like in her research, curiosity, open-mindedness, and a passion for problem-solving have never led her astray. More

  • in

    Startup accelerates progress toward light-speed computing

    Our ability to cram ever-smaller transistors onto a chip has enabled today’s age of ubiquitous computing. But that approach is finally running into limits, with some experts declaring an end to Moore’s Law and a related principle, known as Dennard’s Scaling.

    Those developments couldn’t be coming at a worse time. Demand for computing power has skyrocketed in recent years thanks in large part to the rise of artificial intelligence, and it shows no signs of slowing down.

    Now Lightmatter, a company founded by three MIT alumni, is continuing the remarkable progress of computing by rethinking the lifeblood of the chip. Instead of relying solely on electricity, the company also uses light for data processing and transport. The company’s first two products, a chip specializing in artificial intelligence operations and an interconnect that facilitates data transfer between chips, use both photons and electrons to drive more efficient operations.

    “The two problems we are solving are ‘How do chips talk?’ and ‘How do you do these [AI] calculations?’” Lightmatter co-founder and CEO Nicholas Harris PhD ’17 says. “With our first two products, Envise and Passage, we’re addressing both of those questions.”

    In a nod to the size of the problem and the demand for AI, Lightmatter raised just north of $300 million in 2023 at a valuation of $1.2 billion. Now the company is demonstrating its technology with some of the largest technology companies in the world in hopes of reducing the massive energy demand of data centers and AI models.

    “We’re going to enable platforms on top of our interconnect technology that are made up of hundreds of thousands of next-generation compute units,” Harris says. “That simply wouldn’t be possible without the technology that we’re building.”

    From idea to $100K

    Prior to MIT, Harris worked at the semiconductor company Micron Technology, where he studied the fundamental devices behind integrated chips. The experience made him see how the traditional approach for improving computer performance — cramming more transistors onto each chip — was hitting its limits.

    “I saw how the roadmap for computing was slowing, and I wanted to figure out how I could continue it,” Harris says. “What approaches can augment computers? Quantum computing and photonics were two of those pathways.”

    Harris came to MIT to work on photonic quantum computing for his PhD under Dirk Englund, an associate professor in the Department of Electrical Engineering and Computer Science. As part of that work, he built silicon-based integrated photonic chips that could send and process information using light instead of electricity.

    The work led to dozens of patents and more than 80 research papers in prestigious journals like Nature. But another technology also caught Harris’s attention at MIT.

    “I remember walking down the hall and seeing students just piling out of these auditorium-sized classrooms, watching relayed live videos of lectures to see professors teach deep learning,” Harris recalls, referring to the artificial intelligence technique. “Everybody on campus knew that deep learning was going to be a huge deal, so I started learning more about it, and we realized that the systems I was building for photonic quantum computing could actually be leveraged to do deep learning.”

    Harris had planned to become a professor after his PhD, but he realized he could attract more funding and innovate more quickly through a startup, so he teamed up with Darius Bunandar PhD ’18, who was also studying in Englund’s lab, and Thomas Graham MBA ’18. The co-founders successfully launched into the startup world by winning the 2017 MIT $100K Entrepreneurship Competition.

    Seeing the light

    Lightmatter’s Envise chip takes the part of computing that electrons do well, like memory, and combines it with what light does well, like performing the massive matrix multiplications of deep-learning models.

    “With photonics, you can perform multiple calculations at the same time because the data is coming in on different colors of light,” Harris explains. “In one color, you could have a photo of a dog. In another color, you could have a photo of a cat. In another color, maybe a tree, and you could have all three of those operations going through the same optical computing unit, this matrix accelerator, at the same time. That drives up operations per area, and it reuses the hardware that’s there, driving up energy efficiency.”

    Passage takes advantage of light’s latency and bandwidth advantages to link processors in a manner similar to how fiber optic cables use light to send data over long distances. It also enables chips as big as entire wafers to act as a single processor. Sending information between chips is central to running the massive server farms that power cloud computing and run AI systems like ChatGPT.

    Both products are designed to bring energy efficiencies to computing, which Harris says are needed to keep up with rising demand without bringing huge increases in power consumption.

    “By 2040, some predict that around 80 percent of all energy usage on the planet will be devoted to data centers and computing, and AI is going to be a huge fraction of that,” Harris says. “When you look at computing deployments for training these large AI models, they’re headed toward using hundreds of megawatts. Their power usage is on the scale of cities.”

    Lightmatter is currently working with chipmakers and cloud service providers for mass deployment. Harris notes that because the company’s equipment runs on silicon, it can be produced by existing semiconductor fabrication facilities without massive changes in process.

    The ambitious plans are designed to open up a new path forward for computing that would have huge implications for the environment and economy.

    “We’re going to continue looking at all of the pieces of computers to figure out where light can accelerate them, make them more energy efficient, and faster, and we’re going to continue to replace those parts,” Harris says. “Right now, we’re focused on interconnect with Passage and on compute with Envise. But over time, we’re going to build out the next generation of computers, and it’s all going to be centered around light.” More

  • in

    Power when the sun doesn’t shine

    In 2016, at the huge Houston energy conference CERAWeek, MIT materials scientist Yet-Ming Chiang found himself talking to a Tesla executive about a thorny problem: how to store the output of solar panels and wind turbines for long durations.        

    Chiang, the Kyocera Professor of Materials Science and Engineering, and Mateo Jaramillo, a vice president at Tesla, knew that utilities lacked a cost-effective way to store renewable energy to cover peak levels of demand and to bridge the gaps during windless and cloudy days. They also knew that the scarcity of raw materials used in conventional energy storage devices needed to be addressed if renewables were ever going to displace fossil fuels on the grid at scale.

    Energy storage technologies can facilitate access to renewable energy sources, boost the stability and reliability of power grids, and ultimately accelerate grid decarbonization. The global market for these systems — essentially large batteries — is expected to grow tremendously in the coming years. A study by the nonprofit LDES (Long Duration Energy Storage) Council pegs the long-duration energy storage market at between 80 and 140 terawatt-hours by 2040. “That’s a really big number,” Chiang notes. “Every 10 people on the planet will need access to the equivalent of one EV [electric vehicle] battery to support their energy needs.”

    In 2017, one year after they met in Houston, Chiang and Jaramillo joined forces to co-found Form Energy in Somerville, Massachusetts, with MIT graduates Marco Ferrara SM ’06, PhD ’08 and William Woodford PhD ’13, and energy storage veteran Ted Wiley.

    “There is a burgeoning market for electrical energy storage because we want to achieve decarbonization as fast and as cost-effectively as possible,” says Ferrara, Form’s senior vice president in charge of software and analytics.

    Investors agreed. Over the next six years, Form Energy would raise more than $800 million in venture capital.

    Bridging gaps

    The simplest battery consists of an anode, a cathode, and an electrolyte. During discharge, with the help of the electrolyte, electrons flow from the negative anode to the positive cathode. During charge, external voltage reverses the process. The anode becomes the positive terminal, the cathode becomes the negative terminal, and electrons move back to where they started. Materials used for the anode, cathode, and electrolyte determine the battery’s weight, power, and cost “entitlement,” which is the total cost at the component level.

    During the 1980s and 1990s, the use of lithium revolutionized batteries, making them smaller, lighter, and able to hold a charge for longer. The storage devices Form Energy has devised are rechargeable batteries based on iron, which has several advantages over lithium. A big one is cost.

    Chiang once declared to the MIT Club of Northern California, “I love lithium-ion.” Two of the four MIT spinoffs Chiang founded center on innovative lithium-ion batteries. But at hundreds of dollars a kilowatt-hour (kWh) and with a storage capacity typically measured in hours, lithium-ion was ill-suited for the use he now had in mind.

    The approach Chiang envisioned had to be cost-effective enough to boost the attractiveness of renewables. Making solar and wind energy reliable enough for millions of customers meant storing it long enough to fill the gaps created by extreme weather conditions, grid outages, and when there is a lull in the wind or a few days of clouds.

    To be competitive with legacy power plants, Chiang’s method had to come in at around $20 per kilowatt-hour of stored energy — one-tenth the cost of lithium-ion battery storage.

    But how to transition from expensive batteries that store and discharge over a couple of hours to some as-yet-undefined, cheap, longer-duration technology?

    “One big ball of iron”

    That’s where Ferrara comes in. Ferrara has a PhD in nuclear engineering from MIT and a PhD in electrical engineering and computer science from the University of L’Aquila in his native Italy. In 2017, as a research affiliate at the MIT Department of Materials Science and Engineering, he worked with Chiang to model the grid’s need to manage renewables’ intermittency.

    How intermittent depends on where you are. In the United States, for instance, there’s the windy Great Plains; the sun-drenched, relatively low-wind deserts of Arizona, New Mexico, and Nevada; and the often-cloudy Pacific Northwest.

    Ferrara, in collaboration with Professor Jessika Trancik of MIT’s Institute for Data, Systems, and Society and her MIT team, modeled four representative locations in the United States and concluded that energy storage with capacity costs below roughly $20/kWh and discharge durations of multiple days would allow a wind-solar mix to provide cost-competitive, firm electricity in resource-abundant locations.

    Now that they had a time frame, they turned their attention to materials. At the price point Form Energy was aiming for, lithium was out of the question. Chiang looked at plentiful and cheap sulfur. But a sulfur, sodium, water, and air battery had technical challenges.

    Thomas Edison once used iron as an electrode, and iron-air batteries were first studied in the 1960s. They were too heavy to make good transportation batteries. But this time, Chiang and team were looking at a battery that sat on the ground, so weight didn’t matter. Their priorities were cost and availability.

    “Iron is produced, mined, and processed on every continent,” Chiang says. “The Earth is one big ball of iron. We wouldn’t ever have to worry about even the most ambitious projections of how much storage that the world might use by mid-century.” If Form ever moves into the residential market, “it’ll be the safest battery you’ve ever parked at your house,” Chiang laughs. “Just iron, air, and water.”

    Scientists call it reversible rusting. While discharging, the battery takes in oxygen and converts iron to rust. Applying an electrical current converts the rusty pellets back to iron, and the battery “breathes out” oxygen as it charges. “In chemical terms, you have iron, and it becomes iron hydroxide,” Chiang says. “That means electrons were extracted. You get those electrons to go through the external circuit, and now you have a battery.”

    Form Energy’s battery modules are approximately the size of a washer-and-dryer unit. They are stacked in 40-foot containers, and several containers are electrically connected with power conversion systems to build storage plants that can cover several acres.

    The right place at the right time

    The modules don’t look or act like anything utilities have contracted for before.

    That’s one of Form’s key challenges. “There is not widespread knowledge of needing these new tools for decarbonized grids,” Ferrara says. “That’s not the way utilities have typically planned. They’re looking at all the tools in the toolkit that exist today, which may not contemplate a multi-day energy storage asset.”

    Form Energy’s customers are largely traditional power companies seeking to expand their portfolios of renewable electricity. Some are in the process of decommissioning coal plants and shifting to renewables.

    Ferrara’s research pinpointing the need for very low-cost multi-day storage provides key data for power suppliers seeking to determine the most cost-effective way to integrate more renewable energy.

    Using the same modeling techniques, Ferrara and team show potential customers how the technology fits in with their existing system, how it competes with other technologies, and how, in some cases, it can operate synergistically with other storage technologies.

    “They may need a portfolio of storage technologies to fully balance renewables on different timescales of intermittency,” he says. But other than the technology developed at Form, “there isn’t much out there, certainly not within the cost entitlement of what we’re bringing to market.”  Thanks to Chiang and Jaramillo’s chance encounter in Houston, Form has a several-year lead on other companies working to address this challenge. 

    In June 2023, Form Energy closed its biggest deal to date for a single project: Georgia Power’s order for a 15-megawatt/1,500-megawatt-hour system. That order brings Form’s total amount of energy storage under contracts with utility customers to 40 megawatts/4 gigawatt-hours. To meet the demand, Form is building a new commercial-scale battery manufacturing facility in West Virginia.

    The fact that Form Energy is creating jobs in an area that lost more than 10,000 steel jobs over the past decade is not lost on Chiang. “And these new jobs are in clean tech. It’s super exciting to me personally to be doing something that benefits communities outside of our traditional technology centers.

    “This is the right time for so many reasons,” Chiang says. He says he and his Form Energy co-founders feel “tremendous urgency to get these batteries out into the world.”

    This article appears in the Winter 2024 issue of Energy Futures, the magazine of the MIT Energy Initiative. More

  • in

    New AI model could streamline operations in a robotic warehouse

    Hundreds of robots zip back and forth across the floor of a colossal robotic warehouse, grabbing items and delivering them to human workers for packing and shipping. Such warehouses are increasingly becoming part of the supply chain in many industries, from e-commerce to automotive production.

    However, getting 800 robots to and from their destinations efficiently while keeping them from crashing into each other is no easy task. It is such a complex problem that even the best path-finding algorithms struggle to keep up with the breakneck pace of e-commerce or manufacturing. 

    In a sense, these robots are like cars trying to navigate a crowded city center. So, a group of MIT researchers who use AI to mitigate traffic congestion applied ideas from that domain to tackle this problem.

    They built a deep-learning model that encodes important information about the warehouse, including the robots, planned paths, tasks, and obstacles, and uses it to predict the best areas of the warehouse to decongest to improve overall efficiency.

    Their technique divides the warehouse robots into groups, so these smaller groups of robots can be decongested faster with traditional algorithms used to coordinate robots. In the end, their method decongests the robots nearly four times faster than a strong random search method.

    In addition to streamlining warehouse operations, this deep learning approach could be used in other complex planning tasks, like computer chip design or pipe routing in large buildings.

    “We devised a new neural network architecture that is actually suitable for real-time operations at the scale and complexity of these warehouses. It can encode hundreds of robots in terms of their trajectories, origins, destinations, and relationships with other robots, and it can do this in an efficient manner that reuses computation across groups of robots,” says Cathy Wu, the Gilbert W. Winslow Career Development Assistant Professor in Civil and Environmental Engineering (CEE), and a member of a member of the Laboratory for Information and Decision Systems (LIDS) and the Institute for Data, Systems, and Society (IDSS).

    Wu, senior author of a paper on this technique, is joined by lead author Zhongxia Yan, a graduate student in electrical engineering and computer science. The work will be presented at the International Conference on Learning Representations.

    Robotic Tetris

    From a bird’s eye view, the floor of a robotic e-commerce warehouse looks a bit like a fast-paced game of “Tetris.”

    When a customer order comes in, a robot travels to an area of the warehouse, grabs the shelf that holds the requested item, and delivers it to a human operator who picks and packs the item. Hundreds of robots do this simultaneously, and if two robots’ paths conflict as they cross the massive warehouse, they might crash.

    Traditional search-based algorithms avoid potential crashes by keeping one robot on its course and replanning a trajectory for the other. But with so many robots and potential collisions, the problem quickly grows exponentially.

    “Because the warehouse is operating online, the robots are replanned about every 100 milliseconds. That means that every second, a robot is replanned 10 times. So, these operations need to be very fast,” Wu says.

    Because time is so critical during replanning, the MIT researchers use machine learning to focus the replanning on the most actionable areas of congestion — where there exists the most potential to reduce the total travel time of robots.

    Wu and Yan built a neural network architecture that considers smaller groups of robots at the same time. For instance, in a warehouse with 800 robots, the network might cut the warehouse floor into smaller groups that contain 40 robots each.

    Then, it predicts which group has the most potential to improve the overall solution if a search-based solver were used to coordinate trajectories of robots in that group.

    An iterative process, the overall algorithm picks the most promising robot group with the neural network, decongests the group with the search-based solver, then picks the next most promising group with the neural network, and so on.

    Considering relationships

    The neural network can reason about groups of robots efficiently because it captures complicated relationships that exist between individual robots. For example, even though one robot may be far away from another initially, their paths could still cross during their trips.

    The technique also streamlines computation by encoding constraints only once, rather than repeating the process for each subproblem. For instance, in a warehouse with 800 robots, decongesting a group of 40 robots requires holding the other 760 robots as constraints. Other approaches require reasoning about all 800 robots once per group in each iteration.

    Instead, the researchers’ approach only requires reasoning about the 800 robots once across all groups in each iteration.

    “The warehouse is one big setting, so a lot of these robot groups will have some shared aspects of the larger problem. We designed our architecture to make use of this common information,” she adds.

    They tested their technique in several simulated environments, including some set up like warehouses, some with random obstacles, and even maze-like settings that emulate building interiors.

    By identifying more effective groups to decongest, their learning-based approach decongests the warehouse up to four times faster than strong, non-learning-based approaches. Even when they factored in the additional computational overhead of running the neural network, their approach still solved the problem 3.5 times faster.

    In the future, the researchers want to derive simple, rule-based insights from their neural model, since the decisions of the neural network can be opaque and difficult to interpret. Simpler, rule-based methods could also be easier to implement and maintain in actual robotic warehouse settings.

    “This approach is based on a novel architecture where convolution and attention mechanisms interact effectively and efficiently. Impressively, this leads to being able to take into account the spatiotemporal component of the constructed paths without the need of problem-specific feature engineering. The results are outstanding: Not only is it possible to improve on state-of-the-art large neighborhood search methods in terms of quality of the solution and speed, but the model generalizes to unseen cases wonderfully,” says Andrea Lodi, the Andrew H. and Ann R. Tisch Professor at Cornell Tech, and who was not involved with this research.

    This work was supported by Amazon and the MIT Amazon Science Hub. More

  • in

    Automated method helps researchers quantify uncertainty in their predictions

    Pollsters trying to predict presidential election results and physicists searching for distant exoplanets have at least one thing in common: They often use a tried-and-true scientific technique called Bayesian inference.

    Bayesian inference allows these scientists to effectively estimate some unknown parameter — like the winner of an election — from data such as poll results. But Bayesian inference can be slow, sometimes consuming weeks or even months of computation time or requiring a researcher to spend hours deriving tedious equations by hand. 

    Researchers from MIT and elsewhere have introduced an optimization technique that speeds things up without requiring a scientist to do a lot of additional work. Their method can achieve more accurate results faster than another popular approach for accelerating Bayesian inference.

    Using this new automated technique, a scientist could simply input their model and then the optimization method does all the calculations under the hood to provide an approximation of some unknown parameter. The method also offers reliable uncertainty estimates that can help a researcher understand when to trust its predictions.

    This versatile technique could be applied to a wide array of scientific quandaries that incorporate Bayesian inference. For instance, it could be used by economists studying the impact of microcredit loans in developing nations or sports analysts using a model to rank top tennis players.

    “When you actually dig into what people are doing in the social sciences, physics, chemistry, or biology, they are often using a lot of the same tools under the hood. There are so many Bayesian analyses out there. If we can build a really great tool that makes these researchers lives easier, then we can really make a difference to a lot of people in many different research areas,” says senior author Tamara Broderick, an associate professor in MIT’s Department of Electrical Engineering and Computer Science (EECS) and a member of the Laboratory for Information and Decision Systems and the Institute for Data, Systems, and Society.

    Broderick is joined on the paper by co-lead authors Ryan Giordano, an assistant professor of statistics at the University of California at Berkeley; and Martin Ingram, a data scientist at the AI company KONUX. The paper was recently published in the Journal of Machine Learning Research.

    Faster results

    When researchers seek a faster form of Bayesian inference, they often turn to a technique called automatic differentiation variational inference (ADVI), which is often both fast to run and easy to use.

    But Broderick and her collaborators have found a number of practical issues with ADVI. It has to solve an optimization problem and can do so only approximately. So, ADVI can still require a lot of computation time and user effort to determine whether the approximate solution is good enough. And once it arrives at a solution, it tends to provide poor uncertainty estimates.

    Rather than reinventing the wheel, the team took many ideas from ADVI but turned them around to create a technique called deterministic ADVI (DADVI) that doesn’t have these downsides.

    With DADVI, it is very clear when the optimization is finished, so a user won’t need to spend extra computation time to ensure that the best solution has been found. DADVI also permits the incorporation of more powerful optimization methods that give it an additional speed and performance boost.

    Once it reaches a result, DADVI is set up to allow the use of uncertainty corrections. These corrections make its uncertainty estimates much more accurate than those of ADVI.

    DADVI also enables the user to clearly see how much error they have incurred in the approximation to the optimization problem. This prevents a user from needlessly running the optimization again and again with more and more resources to try and reduce the error.

    “We wanted to see if we could live up to the promise of black-box inference in the sense of, once the user makes their model, they can just run Bayesian inference and don’t have to derive everything by hand, they don’t need to figure out when to stop their algorithm, and they have a sense of how accurate their approximate solution is,” Broderick says.

    Defying conventional wisdom

    DADVI can be more effective than ADVI because it uses an efficient approximation method, called sample average approximation, which estimates an unknown quantity by taking a series of exact steps.

    Because the steps along the way are exact, it is clear when the objective has been reached. Plus, getting to that objective typically requires fewer steps.

    Often, researchers expect sample average approximation to be more computationally intensive than a more popular method, known as stochastic gradient, which is used by ADVI. But Broderick and her collaborators showed that, in many applications, this is not the case.

    “A lot of problems really do have special structure, and you can be so much more efficient and get better performance by taking advantage of that special structure. That is something we have really seen in this paper,” she adds.

    They tested DADVI on a number of real-world models and datasets, including a model used by economists to evaluate the effectiveness of microcredit loans and one used in ecology to determine whether a species is present at a particular site.

    Across the board, they found that DADVI can estimate unknown parameters faster and more reliably than other methods, and achieves as good or better accuracy than ADVI. Because it is easier to use than other techniques, DADVI could offer a boost to scientists in a wide variety of fields.

    In the future, the researchers want to dig deeper into correction methods for uncertainty estimates so they can better understand why these corrections can produce such accurate uncertainties, and when they could fall short.

    “In applied statistics, we often have to use approximate algorithms for problems that are too complex or high-dimensional to allow exact solutions to be computed in reasonable time. This new paper offers an interesting set of theory and empirical results that point to an improvement in a popular existing approximate algorithm for Bayesian inference,” says Andrew Gelman ’85, ’86, a professor of statistics and political science at Columbia University, who was not involved with the study. “As one of the team involved in the creation of that earlier work, I’m happy to see our algorithm superseded by something more stable.”

    This research was supported by a National Science Foundation CAREER Award and the U.S. Office of Naval Research.  More

  • in

    MIT researchers remotely map crops, field by field

    Crop maps help scientists and policymakers track global food supplies and estimate how they might shift with climate change and growing populations. But getting accurate maps of the types of crops that are grown from farm to farm often requires on-the-ground surveys that only a handful of countries have the resources to maintain.

    Now, MIT engineers have developed a method to quickly and accurately label and map crop types without requiring in-person assessments of every single farm. The team’s method uses a combination of Google Street View images, machine learning, and satellite data to automatically determine the crops grown throughout a region, from one fraction of an acre to the next. 

    The researchers used the technique to automatically generate the first nationwide crop map of Thailand — a smallholder country where small, independent farms make up the predominant form of agriculture. The team created a border-to-border map of Thailand’s four major crops — rice, cassava, sugarcane, and maize — and determined which of the four types was grown, at every 10 meters, and without gaps, across the entire country. The resulting map achieved an accuracy of 93 percent, which the researchers say is comparable to on-the-ground mapping efforts in high-income, big-farm countries.

    The team is applying their mapping technique to other countries such as India, where small farms sustain most of the population but the type of crops grown from farm to farm has historically been poorly recorded.

    “It’s a longstanding gap in knowledge about what is grown around the world,” says Sherrie Wang, the d’Arbeloff Career Development Assistant Professor in MIT’s Department of Mechanical Engineering, and the Institute for Data, Systems, and Society (IDSS). “The final goal is to understand agricultural outcomes like yield, and how to farm more sustainably. One of the key preliminary steps is to map what is even being grown — the more granularly you can map, the more questions you can answer.”

    Wang, along with MIT graduate student Jordi Laguarta Soler and Thomas Friedel of the agtech company PEAT GmbH, will present a paper detailing their mapping method later this month at the AAAI Conference on Artificial Intelligence.

    Ground truth

    Smallholder farms are often run by a single family or farmer, who subsist on the crops and livestock that they raise. It’s estimated that smallholder farms support two-thirds of the world’s rural population and produce 80 percent of the world’s food. Keeping tabs on what is grown and where is essential to tracking and forecasting food supplies around the world. But the majority of these small farms are in low to middle-income countries, where few resources are devoted to keeping track of individual farms’ crop types and yields.

    Crop mapping efforts are mainly carried out in high-income regions such as the United States and Europe, where government agricultural agencies oversee crop surveys and send assessors to farms to label crops from field to field. These “ground truth” labels are then fed into machine-learning models that make connections between the ground labels of actual crops and satellite signals of the same fields. They then label and map wider swaths of farmland that assessors don’t cover but that satellites automatically do.

    “What’s lacking in low- and middle-income countries is this ground label that we can associate with satellite signals,” Laguarta Soler says. “Getting these ground truths to train a model in the first place has been limited in most of the world.”

    The team realized that, while many developing countries do not have the resources to maintain crop surveys, they could potentially use another source of ground data: roadside imagery, captured by services such as Google Street View and Mapillary, which send cars throughout a region to take continuous 360-degree images with dashcams and rooftop cameras.

    In recent years, such services have been able to access low- and middle-income countries. While the goal of these services is not specifically to capture images of crops, the MIT team saw that they could search the roadside images to identify crops.

    Cropped image

    In their new study, the researchers worked with Google Street View (GSV) images taken throughout Thailand — a country that the service has recently imaged fairly thoroughly, and which consists predominantly of smallholder farms.

    Starting with over 200,000 GSV images randomly sampled across Thailand, the team filtered out images that depicted buildings, trees, and general vegetation. About 81,000 images were crop-related. They set aside 2,000 of these, which they sent to an agronomist, who determined and labeled each crop type by eye. They then trained a convolutional neural network to automatically generate crop labels for the other 79,000 images, using various training methods, including iNaturalist — a web-based crowdsourced  biodiversity database, and GPT-4V, a “multimodal large language model” that enables a user to input an image and ask the model to identify what the image is depicting. For each of the 81,000 images, the model generated a label of one of four crops that the image was likely depicting — rice, maize, sugarcane, or cassava.

    The researchers then paired each labeled image with the corresponding satellite data taken of the same location throughout a single growing season. These satellite data include measurements across multiple wavelengths, such as a location’s greenness and its reflectivity (which can be a sign of water). 

    “Each type of crop has a certain signature across these different bands, which changes throughout a growing season,” Laguarta Soler notes.

    The team trained a second model to make associations between a location’s satellite data and its corresponding crop label. They then used this model to process satellite data taken of the rest of the country, where crop labels were not generated or available. From the associations that the model learned, it then assigned crop labels across Thailand, generating a country-wide map of crop types, at a resolution of 10 square meters.

    This first-of-its-kind crop map included locations corresponding to the 2,000 GSV images that the researchers originally set aside, that were labeled by arborists. These human-labeled images were used to validate the map’s labels, and when the team looked to see whether the map’s labels matched the expert, “gold standard” labels, it did so 93 percent of the time.

    “In the U.S., we’re also looking at over 90 percent accuracy, whereas with previous work in India, we’ve only seen 75 percent because ground labels are limited,” Wang says. “Now we can create these labels in a cheap and automated way.”

    The researchers are moving to map crops across India, where roadside images via Google Street View and other services have recently become available.

    “There are over 150 million smallholder farmers in India,” Wang says. “India is covered in agriculture, almost wall-to-wall farms, but very small farms, and historically it’s been very difficult to create maps of India because there are very sparse ground labels.”

    The team is working to generate crop maps in India, which could be used to inform policies having to do with assessing and bolstering yields, as global temperatures and populations rise.

    “What would be interesting would be to create these maps over time,” Wang says. “Then you could start to see trends, and we can try to relate those things to anything like changes in climate and policies.” More

  • in

    Six MIT students selected as spring 2024 MIT-Pillar AI Collective Fellows

    The MIT-Pillar AI Collective has announced six fellows for the spring 2024 semester. With support from the program, the graduate students, who are in their final year of a master’s or PhD program, will conduct research in the areas of AI, machine learning, and data science with the aim of commercializing their innovations.

    Launched by MIT’s School of Engineering and Pillar VC in 2022, the MIT-Pillar AI Collective supports faculty, postdocs, and students conducting research on AI, machine learning, and data science. Supported by a gift from Pillar VC and administered by the MIT Deshpande Center for Technological Innovation, the mission of the program is to advance research toward commercialization.

    The spring 2024 MIT-Pillar AI Collective Fellows are:

    Yasmeen AlFaraj

    Yasmeen AlFaraj is a PhD candidate in chemistry whose interest is in the application of data science and machine learning to soft materials design to enable next-generation, sustainable plastics, rubber, and composite materials. More specifically, she is applying machine learning to the design of novel molecular additives to enable the low-cost manufacturing of chemically deconstructable thermosets and composites. AlFaraj’s work has led to the discovery of scalable, translatable new materials that could address thermoset plastic waste. As a Pillar Fellow, she will pursue bringing this technology to market, initially focusing on wind turbine blade manufacturing and conformal coatings. Through the Deshpande Center for Technological Innovation, AlFaraj serves as a lead for a team developing a spinout focused on recyclable versions of existing high-performance thermosets by incorporating small quantities of a degradable co-monomer. In addition, she participated in the National Science Foundation Innovation Corps program and recently graduated from the Clean Tech Open, where she focused on enhancing her business plan, analyzing potential markets, ensuring a complete IP portfolio, and connecting with potential funders. AlFaraj earned a BS in chemistry from University of California at Berkeley.

    Ruben Castro Ornelas

    Ruben Castro Ornelas is a PhD student in mechanical engineering who is passionate about the future of multipurpose robots and designing the hardware to use them with AI control solutions. Combining his expertise in programming, embedded systems, machine design, reinforcement learning, and AI, he designed a dexterous robotic hand capable of carrying out useful everyday tasks without sacrificing size, durability, complexity, or simulatability. Ornelas’s innovative design holds significant commercial potential in domestic, industrial, and health-care applications because it could be adapted to hold everything from kitchenware to delicate objects. As a Pillar Fellow, he will focus on identifying potential commercial markets, determining the optimal approach for business-to-business sales, and identifying critical advisors. Ornelas served as co-director of StartLabs, an undergraduate entrepreneurship club at MIT, where he earned an BS in mechanical engineering.

    Keeley Erhardt

    Keeley Erhardt is a PhD candidate in media arts and sciences whose research interests lie in the transformative potential of AI in network analysis, particularly for entity correlation and hidden link detection within and across domains. She has designed machine learning algorithms to identify and track temporal correlations and hidden signals in large-scale networks, uncovering online influence campaigns originating from multiple countries. She has similarly demonstrated the use of graph neural networks to identify coordinated cryptocurrency accounts by analyzing financial time series data and transaction dynamics. As a Pillar Fellow, Erhardt will pursue the potential commercial applications of her work, such as detecting fraud, propaganda, money laundering, and other covert activity in the finance, energy, and national security sectors. She has had internships at Google, Facebook, and Apple and held software engineering roles at multiple tech unicorns. Erhardt earned an MEng in electrical engineering and computer science and a BS in computer science, both from MIT.

    Vineet Jagadeesan Nair

    Vineet Jagadeesan Nair is a PhD candidate in mechanical engineering whose research focuses on modeling power grids and designing electricity markets to integrate renewables, batteries, and electric vehicles. He is broadly interested in developing computational tools to tackle climate change. As a Pillar Fellow, Nair will explore the application of machine learning and data science to power systems. Specifically, he will experiment with approaches to improve the accuracy of forecasting electricity demand and supply with high spatial-temporal resolution. In collaboration with Project Tapestry @ Google X, he is also working on fusing physics-informed machine learning with conventional numerical methods to increase the speed and accuracy of high-fidelity simulations. Nair’s work could help realize future grids with high penetrations of renewables and other clean, distributed energy resources. Outside academics, Nair is active in entrepreneurship, most recently helping to organize the 2023 MIT Global Startup Workshop in Greece. He earned an MS in computational science and engineering from MIT, an MPhil in energy technologies from Cambridge University as a Gates Scholar, and a BS in mechanical engineering and a BA in economics from University of California at Berkeley.

    Mahdi Ramadan

    Mahdi Ramadan is a PhD candidate in brain and cognitive sciences whose research interests lie at the intersection of cognitive science, computational modeling, and neural technologies. His work uses novel unsupervised methods for learning and generating interpretable representations of neural dynamics, capitalizing on recent advances in AI, specifically contrastive and geometric deep learning techniques capable of uncovering the latent dynamics underlying neural processes with high fidelity. As a Pillar Fellow, he will leverage these methods to gain a better understanding of dynamical models of muscle signals for generative motor control. By supplementing current spinal prosthetics with generative AI motor models that can streamline, speed up, and correct limb muscle activations in real time, as well as potentially using multimodal vision-language models to infer the patients’ high-level intentions, Ramadan aspires to build truly scalable, accessible, and capable commercial neuroprosthetics. Ramadan’s entrepreneurial experience includes being the co-founder of UltraNeuro, a neurotechnology startup, and co-founder of Presizely, a computer vision startup. He earned a BS in neurobiology from University of Washington.

    Rui (Raymond) Zhou

    Rui (Raymond) Zhou is a PhD candidate in mechanical engineering whose research focuses on multimodal AI for engineering design. As a Pillar Fellow, he will advance models that could enable designers to translate information in any modality or combination of modalities into comprehensive 2D and 3D designs, including parametric data, component visuals, assembly graphs, and sketches. These models could also optimize existing human designs to accomplish goals such as improving ergonomics or reducing drag coefficient. Ultimately, Zhou aims to translate his work into a software-as-a-service platform that redefines product design across various sectors, from automotive to consumer electronics. His efforts have the potential to not only accelerate the design process but also reduce costs, opening the door to unprecedented levels of customization, idea generation, and rapid prototyping. Beyond his academic pursuits, Zhou founded UrsaTech, a startup that integrates AI into education and engineering design. He earned a BS in electrical engineering and computer sciences from University of California at Berkeley. More

  • in

    How symmetry can come to the aid of machine learning

    Behrooz Tahmasebi — an MIT PhD student in the Department of Electrical Engineering and Computer Science (EECS) and an affiliate of the Computer Science and Artificial Intelligence Laboratory (CSAIL) — was taking a mathematics course on differential equations in late 2021 when a glimmer of inspiration struck. In that class, he learned for the first time about Weyl’s law, which had been formulated 110 years earlier by the German mathematician Hermann Weyl. Tahmasebi realized it might have some relevance to the computer science problem he was then wrestling with, even though the connection appeared — on the surface — to be thin, at best. Weyl’s law, he says, provides a formula that measures the complexity of the spectral information, or data, contained within the fundamental frequencies of a drum head or guitar string.

    Tahmasebi was, at the same time, thinking about measuring the complexity of the input data to a neural network, wondering whether that complexity could be reduced by taking into account some of the symmetries inherent to the dataset. Such a reduction, in turn, could facilitate — as well as speed up — machine learning processes.

    Weyl’s law, conceived about a century before the boom in machine learning, had traditionally been applied to very different physical situations — such as those concerning the vibrations of a string or the spectrum of electromagnetic (black-body) radiation given off by a heated object. Nevertheless, Tahmasebi believed that a customized version of that law might help with the machine learning problem he was pursuing. And if the approach panned out, the payoff could be considerable.

    He spoke with his advisor, Stefanie Jegelka — an associate professor in EECS and affiliate of CSAIL and the MIT Institute for Data, Systems, and Society — who believed the idea was definitely worth looking into. As Tahmasebi saw it, Weyl’s law had to do with gauging the complexity of data, and so did this project. But Weyl’s law, in its original form, said nothing about symmetry.

    He and Jegelka have now succeeded in modifying Weyl’s law so that symmetry can be factored into the assessment of a dataset’s complexity. “To the best of my knowledge,” Tahmasebi says, “this is the first time Weyl’s law has been used to determine how machine learning can be enhanced by symmetry.”

    The paper he and Jegelka wrote earned a “Spotlight” designation when it was presented at the December 2023 conference on Neural Information Processing Systems — widely regarded as the world’s top conference on machine learning.

    This work, comments Soledad Villar, an applied mathematician at Johns Hopkins University, “shows that models that satisfy the symmetries of the problem are not only correct but also can produce predictions with smaller errors, using a small amount of training points. [This] is especially important in scientific domains, like computational chemistry, where training data can be scarce.”

    In their paper, Tahmasebi and Jegelka explored the ways in which symmetries, or so-called “invariances,” could benefit machine learning. Suppose, for example, the goal of a particular computer run is to pick out every image that contains the numeral 3. That task can be a lot easier, and go a lot quicker, if the algorithm can identify the 3 regardless of where it is placed in the box — whether it’s exactly in the center or off to the side — and whether it is pointed right-side up, upside down, or oriented at a random angle. An algorithm equipped with the latter capability can take advantage of the symmetries of translation and rotations, meaning that a 3, or any other object, is not changed in itself by altering its position or by rotating it around an arbitrary axis. It is said to be invariant to those shifts. The same logic can be applied to algorithms charged with identifying dogs or cats. A dog is a dog is a dog, one might say, irrespective of how it is embedded within an image. 

    The point of the entire exercise, the authors explain, is to exploit a dataset’s intrinsic symmetries in order to reduce the complexity of machine learning tasks. That, in turn, can lead to a reduction in the amount of data needed for learning. Concretely, the new work answers the question: How many fewer data are needed to train a machine learning model if the data contain symmetries?

    There are two ways of achieving a gain, or benefit, by capitalizing on the symmetries present. The first has to do with the size of the sample to be looked at. Let’s imagine that you are charged, for instance, with analyzing an image that has mirror symmetry — the right side being an exact replica, or mirror image, of the left. In that case, you don’t have to look at every pixel; you can get all the information you need from half of the image — a factor of two improvement. If, on the other hand, the image can be partitioned into 10 identical parts, you can get a factor of 10 improvement. This kind of boosting effect is linear.

    To take another example, imagine you are sifting through a dataset, trying to find sequences of blocks that have seven different colors — black, blue, green, purple, red, white, and yellow. Your job becomes much easier if you don’t care about the order in which the blocks are arranged. If the order mattered, there would be 5,040 different combinations to look for. But if all you care about are sequences of blocks in which all seven colors appear, then you have reduced the number of things — or sequences — you are searching for from 5,040 to just one.

    Tahmasebi and Jegelka discovered that it is possible to achieve a different kind of gain — one that is exponential — that can be reaped for symmetries that operate over many dimensions. This advantage is related to the notion that the complexity of a learning task grows exponentially with the dimensionality of the data space. Making use of a multidimensional symmetry can therefore yield a disproportionately large return. “This is a new contribution that is basically telling us that symmetries of higher dimension are more important because they can give us an exponential gain,” Tahmasebi says. 

    The NeurIPS 2023 paper that he wrote with Jegelka contains two theorems that were proved mathematically. “The first theorem shows that an improvement in sample complexity is achievable with the general algorithm we provide,” Tahmasebi says. The second theorem complements the first, he added, “showing that this is the best possible gain you can get; nothing else is achievable.”

    He and Jegelka have provided a formula that predicts the gain one can obtain from a particular symmetry in a given application. A virtue of this formula is its generality, Tahmasebi notes. “It works for any symmetry and any input space.” It works not only for symmetries that are known today, but it could also be applied in the future to symmetries that are yet to be discovered. The latter prospect is not too farfetched to consider, given that the search for new symmetries has long been a major thrust in physics. That suggests that, as more symmetries are found, the methodology introduced by Tahmasebi and Jegelka should only get better over time.

    According to Haggai Maron, a computer scientist at Technion (the Israel Institute of Technology) and NVIDIA who was not involved in the work, the approach presented in the paper “diverges substantially from related previous works, adopting a geometric perspective and employing tools from differential geometry. This theoretical contribution lends mathematical support to the emerging subfield of ‘Geometric Deep Learning,’ which has applications in graph learning, 3D data, and more. The paper helps establish a theoretical basis to guide further developments in this rapidly expanding research area.” More