Markus Andrews

More stories

  • 100 Shares189 Views

    in Data Management & Statistics

    Improving accessibility of online graphics for blind users

    The beauty of a nice infographic published alongside a news or magazine story is that it makes numeric data more accessible to the average reader. But for blind and visually impaired users, such graphics often have the opposite effect.

    For visually impaired users — who frequently rely on screen-reading software that speaks words or numbers aloud as the user moves a cursor across the screen — a graphic may be nothing more than a few words of alt text, such as a chart’s title. For instance, a map of the United States displaying population rates by county might have alt text in the HTML that says simply, “A map of the United States with population rates by county.” The data has been buried in an image, making it entirely inaccessible.

    “Charts have these various visual features that, as a [sighted] reader, you can shift your attention around, look at high-level patterns, look at individual data points, and you can do this on the fly,” says Jonathan Zong, a 2022 MIT Morningside Academy for Design (MAD) Fellow and PhD student in computer science, who points out that even when a graphic includes alt text that interprets the data, the visually impaired user must accept the findings as presented.

    “If you’re [blind and] using a screen reader, the text description imposes a linear predefined reading order. So, you’re beholden to the decisions that the person who wrote the text made about what information was important to include.”

    While some graphics do include data tables that a screen reader can read, it requires the user to remember all the data from each row and column as they move on to the next one. According to the National Federation of the Blind, Zong says, there are 7 million people living in the United States with visual disabilities, and nearly 97 percent of top-level pages on the internet are not accessible to screen readers. The problem, he points out, is an especially difficult one for blind researchers to get around. Some researchers with visual impairments rely on a sighted collaborator to read and help interpret graphics in peer-reviewed research.

    Working with the Visualization Group at the Computer Science and Artificial Intelligence Lab (CSAIL) on a project led by Associate Professor Arvind Satyanarayan that includes Daniel Hajas, a blind researcher and innovation manager at the Global Disability Innovation Hub in England, Zong and others have written an open-source Javascript software program named Olli that solves this problem when it’s included on a website. Olli is able to go from big-picture analysis of a chart to the finest grain of detail to give the user the ability to select the degree of granularity that interests them.

    “We want to design richer screen-reader experiences for visualization with a hierarchical structure, multiple ways to navigate, and descriptions at varying levels of granularity to provide self-guided, open-ended exploration for the user.”

    Next steps with Olli are incorporating multi-sensory software to integrate text and visuals with sound, such as having a musical note that moves up or down the harmonic scale to indicate the direction of data on a linear graph, and possibly even developing tactile interpretations of data. Like most of the MAD Fellows, Zong integrates his science and engineering skills with design and art to create solutions to real-world problems affecting individuals. He’s been recognized for his work in both the visual arts and computer science. He holds undergraduate degrees in computer science and visual arts with a focus on graphic design from Princeton University, where his research was on the ethics of data collection.

    “The throughline is the idea that design can help us make progress on really tough social and ethical questions,” Zong says, calling software for accessible data visualization an “intellectually rich area for design.” “We’re thinking about ways to translate charts and graphs into text descriptions that can get read aloud as speech, or thinking about other kinds of audio mappings to sonify data, and we’re even exploring some tactile methods to understand data,” he says.

    “I get really excited about design when it’s a way to both create things that are useful to people in everyday life and also make progress on larger conversations about technology and society. I think working in accessibility is a great way to do that.”

    Another problem at the intersection of technology and society is the ethics of taking user data from social media for large-scale studies without the users’ awareness. While working as a summer graduate research fellow at Cornell’s Citizens and Technology Lab, Zong helped create an open-source software called Bartleby that can be used in large anonymous data research studies. After researchers collect data, but before analysis, Bartleby would automatically send an email message to every user whose data was included, alert them to that fact and offer them the choice to review the resulting data table and opt out of the study. Bartleby was honored in the student category of Fast Company’s Innovation by Design Awards for 2022. In November the same year, Forbes magazine named Jonathan Zong in its Forbes 30 Under 30 in Science 2023 list for his work in data visualization accessibility.

    The underlying theme to all Zong’s work is the exploration of autonomy and agency, even in his artwork, which is heavily inclusive of text and semiotic play. In “Public Display,” he created a handmade digital display font by erasing parts of celebrity faces that were taken from a facial recognition dataset. The piece was exhibited in 2020 in MIT’s Wiesner Gallery, and received the third-place prize in the MIT Schnitzer Prize in the Visual Arts that year. The work deals not only with the neurological aspects of distinguishing faces from typefaces, but also with the implications for erasing individuals’ identities through the practice of using facial recognition programs that often target individuals in communities of color in unfair ways. Another of his works, “Biometric Sans,” a typography system that stretches letters based on a person’s typing speed, will be included in a show at the Harvard Science Center sometime next fall.

    “MAD, particularly the large events MAD jointly hosted, played a really important function in showing the rest of MIT that this is the kind of work we value. This is what design can look like and is capable of doing. I think it all contributes to that culture shift where this kind of interdisciplinary work can be valued, recognized, and serve the public.

    “There are shared ideas around embodiment and representation that tie these different pursuits together for me,” Zong says. “In the ethics work, and the art on surveillance, I’m thinking about whether data collectors are representing people the way they want to be seen through data. And similarly, the accessibility work is about whether we can make systems that are flexible to the way people want to use them.” More

  • 163 Shares169 Views

    in Data Management & Statistics

    A more effective experimental design for engineering a cell into a new state

    A strategy for cellular reprogramming involves using targeted genetic interventions to engineer a cell into a new state. The technique holds great promise in immunotherapy, for instance, where researchers could reprogram a patient’s T-cells so they are more potent cancer killers. Someday, the approach could also help identify life-saving cancer treatments or regenerative therapies that repair disease-ravaged organs.

    But the human body has about 20,000 genes, and a genetic perturbation could be on a combination of genes or on any of the over 1,000 transcription factors that regulate the genes. Because the search space is vast and genetic experiments are costly, scientists often struggle to find the ideal perturbation for their particular application.   

    Researchers from MIT and Harvard University developed a new, computational approach that can efficiently identify optimal genetic perturbations based on a much smaller number of experiments than traditional methods.

    Their algorithmic technique leverages the cause-and-effect relationship between factors in a complex system, such as genome regulation, to prioritize the best intervention in each round of sequential experiments.

    The researchers conducted a rigorous theoretical analysis to determine that their technique did, indeed, identify optimal interventions. With that theoretical framework in place, they applied the algorithms to real biological data designed to mimic a cellular reprogramming experiment. Their algorithms were the most efficient and effective.

    “Too often, large-scale experiments are designed empirically. A careful causal framework for sequential experimentation may allow identifying optimal interventions with fewer trials, thereby reducing experimental costs,” says co-senior author Caroline Uhler, a professor in the Department of Electrical Engineering and Computer Science (EECS) who is also co-director of the Eric and Wendy Schmidt Center at the Broad Institute of MIT and Harvard, and a researcher at MIT’s Laboratory for Information and Decision Systems (LIDS) and Institute for Data, Systems and Society (IDSS).

    Joining Uhler on the paper, which appears today in Nature Machine Intelligence, are lead author Jiaqi Zhang, a graduate student and Eric and Wendy Schmidt Center Fellow; co-senior author Themistoklis P. Sapsis, professor of mechanical and ocean engineering at MIT and a member of IDSS; and others at Harvard and MIT.

    Active learning

    When scientists try to design an effective intervention for a complex system, like in cellular reprogramming, they often perform experiments sequentially. Such settings are ideally suited for the use of a machine-learning approach called active learning. Data samples are collected and used to learn a model of the system that incorporates the knowledge gathered so far. From this model, an acquisition function is designed — an equation that evaluates all potential interventions and picks the best one to test in the next trial.

    This process is repeated until an optimal intervention is identified (or resources to fund subsequent experiments run out).

    “While there are several generic acquisition functions to sequentially design experiments, these are not effective for problems of such complexity, leading to very slow convergence,” Sapsis explains.

    Acquisition functions typically consider correlation between factors, such as which genes are co-expressed. But focusing only on correlation ignores the regulatory relationships or causal structure of the system. For instance, a genetic intervention can only affect the expression of downstream genes, but a correlation-based approach would not be able to distinguish between genes that are upstream or downstream.

    “You can learn some of this causal knowledge from the data and use that to design an intervention more efficiently,” Zhang explains.

    The MIT and Harvard researchers leveraged this underlying causal structure for their technique. First, they carefully constructed an algorithm so it can only learn models of the system that account for causal relationships.

    Then the researchers designed the acquisition function so it automatically evaluates interventions using information on these causal relationships. They crafted this function so it prioritizes the most informative interventions, meaning those most likely to lead to the optimal intervention in subsequent experiments.

    “By considering causal models instead of correlation-based models, we can already rule out certain interventions. Then, whenever you get new data, you can learn a more accurate causal model and thereby further shrink the space of interventions,” Uhler explains.

    This smaller search space, coupled with the acquisition function’s special focus on the most informative interventions, is what makes their approach so efficient.

    The researchers further improved their acquisition function using a technique known as output weighting, inspired by the study of extreme events in complex systems. This method carefully emphasizes interventions that are likely to be closer to the optimal intervention.

    “Essentially, we view an optimal intervention as an ‘extreme event’ within the space of all possible, suboptimal interventions and use some of the ideas we have developed for these problems,” Sapsis says.    

    Enhanced efficiency

    They tested their algorithms using real biological data in a simulated cellular reprogramming experiment. For this test, they sought a genetic perturbation that would result in a desired shift in average gene expression. Their acquisition functions consistently identified better interventions than baseline methods through every step in the multi-stage experiment.

    “If you cut the experiment off at any stage, ours would still be more efficient than the baselines. This means you could run fewer experiments and get the same or better results,” Zhang says.

    The researchers are currently working with experimentalists to apply their technique toward cellular reprogramming in the lab.

    Their approach could also be applied to problems outside genomics, such as identifying optimal prices for consumer products or enabling optimal feedback control in fluid mechanics applications.

    In the future, they plan to enhance their technique for optimizations beyond those that seek to match a desired mean. In addition, their method assumes that scientists already understand the causal relationships in their system, but future work could explore how to use AI to learn that information, as well.

    This work was funded, in part, by the Office of Naval Research, the MIT-IBM Watson AI Lab, the MIT J-Clinic for Machine Learning and Health, the Eric and Wendy Schmidt Center at the Broad Institute, a Simons Investigator Award, the Air Force Office of Scientific Research, and a National Science Foundation Graduate Fellowship. More

  • 188 Shares159 Views

    in Data Management & Statistics

    J-PAL North America and Results for America announce 18 collaborations with state and local governments

    J-PAL North America and Results for America have announced 18 new partnerships with state and local governments across the country through their Leveraging Evidence and Evaluation for Equitable Recovery (LEVER) programming, which launched in April of this year. 

    As state and local leaders leverage federal relief funding to invest in their communities, J-PAL North America and Results for America are providing in-depth support to agencies in using data, evaluation, and evidence to advance effective and equitable government programming for generations to come. The 18 new collaborators span the contiguous United States and represent a wide range of pressing and innovative uses of federal Covid-19 recovery funding.

    These partnerships are a key component of the LEVER program, run by J-PAL North America — a regional office of MIT’s Abdul Latif Jameel Poverty Action Lab (J-PAL) — and Results for America — a nonprofit organization that helps government agencies harness the power of evidence and data. Through 2024, LEVER will continue to provide a suite of resources, training, and evaluation design services to prepare state and local government agencies to rigorously evaluate their own programs and to harness existing evidence in developing programs and policies using federal recovery dollars.

    J-PAL North America is working with four leading government agencies following a call for proposals to the LEVER Evaluation Incubator in June. These agencies will work with J-PAL staff to design randomized evaluations to understand the causal impact of important programs that contribute to their jurisdictions’ recovery from Covid-19.

    Connecticut’s Medicaid office, operating out of the state’s Department of Social Services, is working to improve vaccine access and awareness among youth. “Connecticut Medicaid is thrilled to work with J-PAL North America. The technical expertise and training that we receive will expand our knowledge during ‘testing and learning’ interventions that improve the health of our members,” says Gui Woolston, the director of Medicaid and Division of Health Services. 

    Athens-Clarke County Unified Government is invested in evaluating programming for youth development and violence prevention implemented by the Boys and Girls Club of Athens. Their goal is “to measure and transparently communicate program impact,” explains Paige Seago, the data and outcomes coordinator for the American Rescue Plan Act. “The ability to continually iterate and tailor programs to better meet community goals is crucial to long-term success.”

    The County of San Diego’s newly formed Office of Evaluation, Performance, and Analytics is evaluating a pilot program providing rental subsidies for older adults. “Randomized evaluation can help us understand if rent subsidies will help prevent seniors from becoming homeless and will give us useful information about how to move forward,” says Chief Evaluation Officer Ricardo Basurto-Dávila. 

    In King County, Washington, the Executive Climate Office is planning to evaluate efforts to increase equitable access to household energy efficiency programs. “Because of J-PAL’s support, we have confidence that we can reduce climate impacts and extend home electrification benefits to lower-income homeowners in King County — homeowners who otherwise may not have the ability to participate in the clean energy transition,” says King County Climate Director Marissa Aho.

    Fourteen additional state and local agencies are working with Results for America as part of the LEVER Training Sprint. Together, they will develop policies that catalyze sustainable evidence building within government. 

    Jurisdictions selected for the Training Sprint represent government leaders at the city, county, and state levels — all of whom are committed to creating an evaluation framework for policy that will prioritize evidence-based decision-making across the country. Over the course of 10 weeks, with access to tools and coaching, each team will develop an internal implementation policy by embedding key evaluation and evidence practices into their jurisdiction’s decision-making processes. Participants will finish the Training Sprint with a robust decision-making framework that translates their LEVER implementation policies into actionable planning guidance. 

    Government leaders will utilize the LEVER Training Sprint to build a culture of data and evidence focused on leveraging evaluation policies to invest in delivering tangible results for their residents. About their participation in the LEVER Training Sprint, Dana Williams from Denver, Colorado says, “Impact evaluation is such an integral piece to understanding the past, present, and future. I’m excited to participate in the LEVER Training Sprint to better inform and drive evidence-based programming in Denver.”

    The Training Sprint is a part of a growing movement to ground government innovation in data and evidence. Kermina Hanna from the State of New Jersey notes, “It’s vital that we cement a data-driven commitment to equity in government operations, and I’m really excited for this opportunity to develop a national network of colleagues in government who share this passion and dedication to responsive public service.”

    Jurisdictions selected for the Training Sprint are: 

    Boston, Massachusetts;
    Carlsbad, California;
    Connecticut;
    Dallas, Texas;
    Denver City/County, Colorado;
    Fort Collins, Colorado;
    Guilford County, North Carolina;
    King County, Washington;
    Long Beach, California;
    Los Angeles, California;
    New Jersey;
    New Mexico;
    Pittsburgh, Pennsylvania; and
    Washington County, Oregon.
    Those interested in learning more can fill out the LEVER intake form. Please direct any questions about the Evaluation Incubator to Louise Geraghty and questions about the Training Sprint to Chelsea Powell. More

  • 125 Shares179 Views

    in Data Management & Statistics

    MIT welcomes nine MLK Visiting Professors and Scholars for 2023-24

    Established in 1990, the MLK Visiting Professors and Scholars Program at MIT welcomes outstanding scholars to the Institute for visiting appointments. MIT aspires to attract candidates who are, in the words of Martin Luther King Jr., “trailblazers in human, academic, scientific and religious freedom.” The program honors King’s life and legacy by expanding and extending the reach of our community. 

    The MLK Scholars Program has welcomed more than 140 professors, practitioners, and professionals at the forefront of their respective fields to MIT. They contribute to the growth and enrichment of the community through their interactions with students, staff, and faculty. They pay tribute to Martin Luther King Jr.’s life and legacy of service and social justice, and they embody MIT’s values: excellence and curiosity, openness and respect, and belonging and community.  

    Each new cohort of scholars actively participates in community engagement and supports MIT’s mission of “advancing knowledge and educating students in science, technology, and other areas of scholarship that will best serve the nation and the world in the 21st century.” 

    The 2023-2024 MLK Scholars:

    Tawanna Dillahunt is an associate professor at the University of Michigan’s School of Information with a joint appointment in their electrical engineering and computer science department. She is joining MIT at the end of a one-year visiting appointment as a Harvard Radcliffe Fellow. Her faculty hosts at the Institute are Catherine D’Ignazio in the Department of Urban Studies and Planning and Fotini Christia in the Institute for Data, Systems, and Society (IDSS). Dillahunt’s research focuses on equitable and inclusive computing. During her appointment, she will host a podcast to explore ethical and socially responsible ways to engage with communities, with a special emphasis on technology. 

    Kwabena Donkor is an assistant professor of marketing at Stanford Graduate School of Business; he is hosted by Dean Eckles, an associate professor of marketing at MIT Sloan School of Management. Donkor’s work bridges economics, psychology, and marketing. His scholarship combines insights from behavioral economics with data and field experiments to study social norms, identity, and how these constructs interact with policy in the marketplace.

    Denise Frazier joins MIT from Tulane University, where she is an assistant director in the New Orleans Center for the Gulf South. She is a researcher and performer and brings a unique interdisciplinary approach to her work at the intersection of cultural studies, environmental justice, and music. Frazier is hosted by Christine Ortiz, the Morris Cohen Professor in the Department of Materials Science and Engineering. 

    Wasalu Jaco, an accomplished performer and artist, is renewing his appointment at MIT for a second year; he is hosted jointly by Nick Montfort, a professor of digital media in the Comparative Media Studies Program/Writing, and Mary Fuller, a professor in the Literature Section and the current chair of the MIT faculty. In his second year, Jaco will work on Cyber/Cypher Rapper, a research project to develop a computational system that participates in responsive and improvisational rap.

    Morgane Konig first joined the Center for Theoretical Physics at MIT in December 2021 as a postdoc. Now a member of the 2023–24 MLK Visiting Scholars Program cohort, she will deepen her ties with scholars and research groups working in cosmology, primarily on early-universe inflation and late-universe signatures that could enable the scientific community to learn more about the mysterious nature of dark matter and dark energy. Her faculty hosts are David Kaiser, the Germeshausen Professor of the History of Science and professor of physics, and Alan Guth, the Victor F. Weisskopf Professor of Physics, both from the Department of Physics.

    The former minister of culture for Colombia and a transformational leader dedicated to environmental protection, Angelica Mayolo-Obregon joins MIT from Buenaventura, Colombia. During her time at MIT, she will serve as an advisor and guest speaker, and help MIT facilitate gatherings of environmental leaders committed to addressing climate action and conserving biodiversity across the Americas, with a special emphasis on Afro-descendant communities. Mayolo-Obregon is hosted by John Fernandez, a professor of building technology in the Department of Architecture and director of MIT’s Environmental Solutions Initiative, and by J. Phillip Thompson, an associate professor in the Department of Urban Studies and Planning (and a former MLK Scholar).

    Jean-Luc Pierite is a member of the Tunica-Biloxi Tribe of Louisiana and the president of the board of directors of North American Indian Center of Boston. While at MIT, Pierite will build connections between MIT and the local Indigenous communities. His research focuses on enhancing climate resilience planning by infusing Indigenous knowledge and ecological practices into scientific and other disciplines. His faculty host is Janelle Knox-Hayes, the Lister Brothers Professor of Economic Geography and Planning in the Department of Urban Studies and Planning.

    Christine Taylor-Butler ’81 is a children’s book author who has written over 90 books; she is hosted by Graham Jones, an associate professor of anthropology. An advocate for literacy and STEAM education in underserved urban and rural schools, Taylor-Butler will partner with community organizations in the Boston area. She is also completing the fourth installment of her middle-grade series, “The Lost Tribe.” These books follow a team of five kids as they use science and technology to crack codes and solve mysteries.

    Angelino Viceisza, a professor of economics at Spelman College, joins MIT Sloan as an MLK Visiting Professor and the Phyllis Wallace Visiting Professor; he is hosted by Robert Gibbons, Sloan Distinguished Professor of Management, and Ray Reagans, Alfred P. Sloan Professor of Management, professor of organization studies, and associate dean for diversity, equity, and inclusion at MIT Sloan. Viceisza has strong, ongoing connections with MIT. His research focuses on remittances, retirement, and household finance in low-income countries and is relevant to public finance and financial economics, as well as the development and organizational economics communities at MIT. 

    Javit Drake, Moriba Jah, and Louis Massiah, members of last year’s cohort of MLK Scholars, will remain at MIT through the end of 2023.

    There are multiple opportunities throughout the year to meet our MLK Visiting Scholars and learn more about their research projects and their social impact. 

    For more information about the MLK Visiting Professors and Scholars Program and upcoming events, visit the website. More

  • 63 Shares129 Views

    in Data Management & Statistics

    Improving US air quality, equitably

    Decarbonization of national economies will be key to achieving global net-zero emissions by 2050, a major stepping stone to the Paris Agreement’s long-term goal of keeping global warming well below 2 degrees Celsius (and ideally 1.5 C), and thereby averting the worst consequences of climate change. Toward that end, the United States has pledged to reduce its greenhouse gas emissions by 50-52 percent from 2005 levels by 2030, backed by its implementation of the 2022 Inflation Reduction Act. This strategy is consistent with a 50-percent reduction in carbon dioxide (CO2) by the end of the decade.

    If U.S. federal carbon policy is successful, the nation’s overall air quality will also improve. Cutting CO2 emissions reduces atmospheric concentrations of air pollutants that lead to the formation of fine particulate matter (PM2.5), which causes more than 200,000 premature deaths in the United States each year. But an average nationwide improvement in air quality will not be felt equally; air pollution exposure disproportionately harms people of color and lower-income populations.

    How effective are current federal decarbonization policies in reducing U.S. racial and economic disparities in PM2.5 exposure, and what changes will be needed to improve their performance? To answer that question, researchers at MIT and Stanford University recently evaluated a range of policies which, like current U.S. federal carbon policies, reduce economy-wide CO2 emissions by 40-60 percent from 2005 levels by 2030. Their findings appear in an open-access article in the journal Nature Communications.

    First, they show that a carbon-pricing policy, while effective in reducing PM2.5 exposure for all racial/ethnic groups, does not significantly mitigate relative disparities in exposure. On average, the white population undergoes far less exposure than Black, Hispanic, and Asian populations. This policy does little to reduce exposure disparities because the CO2 emissions reductions that it achieves primarily occur in the coal-fired electricity sector. Other sectors, such as industry and heavy-duty diesel transportation, contribute far more PM2.5-related emissions.

    The researchers then examine thousands of different reduction options through an optimization approach to identify whether any possible combination of carbon dioxide reductions in the range of 40-60 percent can mitigate disparities. They find that that no policy scenario aligned with current U.S. carbon dioxide emissions targets is likely to significantly reduce current PM2.5 exposure disparities.

    “Policies that address only about 50 percent of CO2 emissions leave many polluting sources in place, and those that prioritize reductions for minorities tend to benefit the entire population,” says Noelle Selin, supervising author of the study and a professor at MIT’s Institute for Data, Systems and Society and Department of Earth, Atmospheric and Planetary Sciences. “This means that a large range of policies that reduce CO2 can improve air quality overall, but can’t address long-standing inequities in air pollution exposure.”

    So if climate policy alone cannot adequately achieve equitable air quality results, what viable options remain? The researchers suggest that more ambitious carbon policies could narrow racial and economic PM2.5 exposure disparities in the long term, but not within the next decade. To make a near-term difference, they recommend interventions designed to reduce PM2.5 emissions resulting from non-CO2 sources, ideally at the economic sector or community level.

    “Achieving improved PM2.5 exposure for populations that are disproportionately exposed across the United States will require thinking that goes beyond current CO2 policy strategies, most likely involving large-scale structural changes,” says Selin. “This could involve changes in local and regional transportation and housing planning, together with accelerated efforts towards decarbonization.” More

  • 150 Shares199 Views

    in Data Management & Statistics

    From physics to generative AI: An AI model for advanced pattern generation

    Generative AI, which is currently riding a crest of popular discourse, promises a world where the simple transforms into the complex — where a simple distribution evolves into intricate patterns of images, sounds, or text, rendering the artificial startlingly real. 

    The realms of imagination no longer remain as mere abstractions, as researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have brought an innovative AI model to life. Their new technology integrates two seemingly unrelated physical laws that underpin the best-performing generative models to date: diffusion, which typically illustrates the random motion of elements, like heat permeating a room or a gas expanding into space, and Poisson Flow, which draws on the principles governing the activity of electric charges.

    This harmonious blend has resulted in superior performance in generating new images, outpacing existing state-of-the-art models. Since its inception, the “Poisson Flow Generative Model ++” (PFGM++) has found potential applications in various fields, from antibody and RNA sequence generation to audio production and graph generation.

    The model can generate complex patterns, like creating realistic images or mimicking real-world processes. PFGM++ builds off of PFGM, the team’s work from the prior year. PFGM takes inspiration from the means behind the mathematical equation known as the “Poisson” equation, and then applies it to the data the model tries to learn from. To do this, the team used a clever trick: They added an extra dimension to their model’s “space,” kind of like going from a 2D sketch to a 3D model. This extra dimension gives more room for maneuvering, places the data in a larger context, and helps one approach the data from all directions when generating new samples. 

    “PFGM++ is an example of the kinds of AI advances that can be driven through interdisciplinary collaborations between physicists and computer scientists,” says Jesse Thaler, theoretical particle physicist in MIT’s Laboratory for Nuclear Science’s Center for Theoretical Physics and director of the National Science Foundation’s AI Institute for Artificial Intelligence and Fundamental Interactions (NSF AI IAIFI), who was not involved in the work. “In recent years, AI-based generative models have yielded numerous eye-popping results, from photorealistic images to lucid streams of text. Remarkably, some of the most powerful generative models are grounded in time-tested concepts from physics, such as symmetries and thermodynamics. PFGM++ takes a century-old idea from fundamental physics — that there might be extra dimensions of space-time — and turns it into a powerful and robust tool to generate synthetic but realistic datasets. I’m thrilled to see the myriad of ways ‘physics intelligence’ is transforming the field of artificial intelligence.”

    The underlying mechanism of PFGM isn’t as complex as it might sound. The researchers compared the data points to tiny electric charges placed on a flat plane in a dimensionally expanded world. These charges produce an “electric field,” with the charges looking to move upwards along the field lines into an extra dimension and consequently forming a uniform distribution on a vast imaginary hemisphere. The generation process is like rewinding a videotape: starting with a uniformly distributed set of charges on the hemisphere and tracking their journey back to the flat plane along the electric lines, they align to match the original data distribution. This intriguing process allows the neural model to learn the electric field, and generate new data that mirrors the original. 

    The PFGM++ model extends the electric field in PFGM to an intricate, higher-dimensional framework. When you keep expanding these dimensions, something unexpected happens — the model starts resembling another important class of models, the diffusion models. This work is all about finding the right balance. The PFGM and diffusion models sit at opposite ends of a spectrum: one is robust but complex to handle, the other simpler but less sturdy. The PFGM++ model offers a sweet spot, striking a balance between robustness and ease of use. This innovation paves the way for more efficient image and pattern generation, marking a significant step forward in technology. Along with adjustable dimensions, the researchers proposed a new training method that enables more efficient learning of the electric field. 

    To bring this theory to life, the team resolved a pair of differential equations detailing these charges’ motion within the electric field. They evaluated the performance using the Frechet Inception Distance (FID) score, a widely accepted metric that assesses the quality of images generated by the model in comparison to the real ones. PFGM++ further showcases a higher resistance to errors and robustness toward the step size in the differential equations.

    Looking ahead, they aim to refine certain aspects of the model, particularly in systematic ways to identify the “sweet spot” value of D tailored for specific data, architectures, and tasks by analyzing the behavior of estimation errors of neural networks. They also plan to apply the PFGM++ to the modern large-scale text-to-image/text-to-video generation.

    “Diffusion models have become a critical driving force behind the revolution in generative AI,” says Yang Song, research scientist at OpenAI. “PFGM++ presents a powerful generalization of diffusion models, allowing users to generate higher-quality images by improving the robustness of image generation against perturbations and learning errors. Furthermore, PFGM++ uncovers a surprising connection between electrostatics and diffusion models, providing new theoretical insights into diffusion model research.”

    “Poisson Flow Generative Models do not only rely on an elegant physics-inspired formulation based on electrostatics, but they also offer state-of-the-art generative modeling performance in practice,” says NVIDIA Senior Research Scientist Karsten Kreis, who was not involved in the work. “They even outperform the popular diffusion models, which currently dominate the literature. This makes them a very powerful generative modeling tool, and I envision their application in diverse areas, ranging from digital content creation to generative drug discovery. More generally, I believe that the exploration of further physics-inspired generative modeling frameworks holds great promise for the future and that Poisson Flow Generative Models are only the beginning.”

    Authors on a paper about this work include three MIT graduate students: Yilun Xu of the Department of Electrical Engineering and Computer Science (EECS) and CSAIL, Ziming Liu of the Department of Physics and the NSF AI IAIFI, and Shangyuan Tong of EECS and CSAIL, as well as Google Senior Research Scientist Yonglong Tian PhD ’23. MIT professors Max Tegmark and Tommi Jaakkola advised the research.

    The team was supported by the MIT-DSTA Singapore collaboration, the MIT-IBM Grand Challenge project, National Science Foundation grants, The Casey and Family Foundation, the Foundational Questions Institute, the Rothberg Family Fund for Cognitive Science, and the ML for Pharmaceutical Discovery and Synthesis Consortium. Their work was presented at the International Conference on Machine Learning this summer. More

  • 175 Shares189 Views

    in Data Management & Statistics

    3 Questions: A new PhD program from the Center for Computational Science and Engineering

    This fall, the Center for Computational Science and Engineering (CCSE), an academic unit in the MIT Schwarzman College of Computing, is introducing a new standalone PhD degree program that will enable students to pursue research in cross-cutting methodological aspects of computational science and engineering. The launch follows approval of the center’s degree program proposal at the May 2023 Institute faculty meeting.

    Doctoral-level graduate study in computational science and engineering (CSE) at MIT has, for the past decade, been offered through an interdisciplinary program in which CSE students are admitted to one of eight participating academic departments in the School of Engineering or School of Science. While this model adds a strong disciplinary component to students’ education, the rapid growth of the CSE field and the establishment of the MIT Schwarzman College of Computing have prompted an exciting expansion of MIT’s graduate-level offerings in computation.

    The new degree, offered by the college, will run alongside MIT’s existing interdisciplinary offerings in CSE, complementing these doctoral training programs and preparing students to contribute to the leading edge of the field. Here, CCSE co-directors Youssef Marzouk and Nicolas Hadjiconstantinou discuss the standalone program and how they expect it to elevate the visibility and impact of CSE research and education at MIT.

    Q: What is computational science and engineering?

    Marzouk: Computational science and engineering focuses on the development and analysis of state-of-the-art methods for computation and their innovative application to problems of science and engineering interest. It has intellectual foundations in applied mathematics, statistics, and computer science, and touches the full range of science and engineering disciplines. Yet, it synthesizes these foundations into a discipline of its own — one that links the digital and physical worlds. It’s an exciting and evolving multidisciplinary field.

    Hadjiconstantinou: Examples of CSE research happening at MIT include modeling and simulation techniques, the underlying computational mathematics, and data-driven modeling of physical systems. Computational statistics and scientific machine learning have become prominent threads within CSE, joining high-performance computing, mathematically-oriented programming languages, and their broader links to algorithms and software. Application domains include energy, environment and climate, materials, health, transportation, autonomy, and aerospace, among others. Some of our researchers focus on general and widely applicable methodology, while others choose to focus on methods and algorithms motivated by a specific domain of application.

    Q: What was the motivation behind creating a standalone PhD program?

    Marzouk: The new degree focuses on a particular class of students whose background and interests are primarily in CSE methodology, in a manner that cuts across the disciplinary research structure represented by our current “with-departments” degree program. There is a strong research demand for such methodologically-focused students among CCSE faculty and MIT faculty in general. Our objective is to create a targeted, coherent degree program in this field that, alongside our other thriving CSE offerings, will create the leading environment for top CSE students worldwide.

    Hadjiconstantinou: One of CCSE’s most important functions is to recruit exceptional students who are trained in and want to work in computational science and engineering. Experience with our CSE master’s program suggests that students with a strong background and interests in the discipline prefer to apply to a pure CSE program for their graduate studies. The standalone degree aims to bring these students to MIT and make them available to faculty across the Institute.

    Q: How will this impact computing education and research at MIT? 

    Hadjiconstantinou: We believe that offering a standalone PhD program in CSE alongside the existing “with-departments” programs will significantly strengthen MIT’s graduate programs in computing. In particular, it will strengthen the methodological core of CSE research and education at MIT, while continuing to support the disciplinary-flavored CSE work taking place in our participating departments, which include Aeronautics and Astronautics; Chemical Engineering; Civil and Environmental Engineering; Materials Science and Engineering; Mechanical Engineering; Nuclear Science and Engineering; Earth, Atmospheric and Planetary Sciences; and Mathematics. Together, these programs will create a stronger CSE student cohort and facilitate deeper exchanges between the college and other units at MIT.

    Marzouk: In a broader sense, the new program is designed to help realize one of the key opportunities presented by the college, which is to create a richer variety of graduate degrees in computation and to involve as many faculty and units in these educational endeavors as possible. The standalone CSE PhD will join other distinguished doctoral programs of the college — such as the Department of Electrical Engineering and Computer Science PhD; the Operations Research Center PhD; and the Interdisciplinary Doctoral Program in Statistics and the Social and Engineering Systems PhD within the Institute for Data, Systems, and Society — and grow in a way that is informed by them. The confluence of these academic programs, and natural synergies among them, will make MIT quite unique. More

  • 75 Shares169 Views

    in Data Management & Statistics

    On the hunt for sustainable materials

    By the time she started high school, Avni Singhal had attended six different schools in a variety of settings, from a traditional public school to a self-paced program. The transitions opened her eyes to how widely educational environments can vary, and made her think about that impact on students.

    “Experiencing so many different types of educational systems exposed me to different ways of looking at things and how that shapes people’s worldviews,” says Singhal.

    Now a fourth-year PhD student in the Department of Materials Science and Engineering, Singhal is still thinking about increasing opportunities for her fellow students, while also pursuing her research. She devotes herself to both developing sustainable materials and improving the graduate experience in her department.

    She recently completed her two-year term as a student representative on the department’s graduate studies committee. In this role, she helped revamp the communication around the qualifying exams and introducing student input to the faculty search process.

    “It’s given me a lot of insight into how our department works,” says Singhal. “It’s a chance to get to know faculty, bring up issues that students experience, and work on changing things that we think could be improved.”

    At the same time, Singhal uses atomistic simulations to model material properties, with an eye toward sustainability. She is a part of the Learning Matter Lab, a group that merges data science tools with engineering and physics-based simulation to better design and understand materials. As part of a computational group, Singhal has worked on a range of projects in collaboration with other labs that are looking to combine computing with other disciplines. Some of this work is sponsored by the MIT Climate and Sustainability Consortium, which facilitates connections across MIT labs and industry.

    Joining the Learning Matter Lab was a step out of Singhal’s comfort zone. She arrived at MIT from the University of California at Berkeley with a joint degree in materials science and bioengineering, as well as a degree in electrical engineering and computer science.

    “I was generally interested in doing work on environment-related applications,” says Singhal. “I was pretty hesitant at first to switch entirely to computation because it’s a very different type of lifestyle of research than what I was doing before.”

    Singhal has taken the challenge in stride, contributing to projects including improving carbon capture molecules and developing new deconstructable, degradable plastics. Not only does Singhal have to understand the technical details of her own work, she also needs to understand the big picture and how to best wield the expertise of her collaborators.

    “When I came in, I was very wide-eyed, thinking computation can do everything because I had never done it before,” says Singhal. “It’s that curve where you know a little bit about something, and you think it can do everything. And then as you learn more, you learn where it can and can’t help us, where it can be valuable, and how to figure out in what part of a project it’s useful.”

    Singhal applies a similarly critical lens when thinking about graduate school as a whole. She notes that access to information and resources is often the main factor determining who enters selective educational programs, and that such access becomes increasingly limited at the graduate level.

    “I realized just how much applying is a function of knowing how to do it,” says Singhal, who co-organized and volunteers with the DMSE Application Assistance Program. The program matches prospective applicants with current students to give feedback on their application materials and provide insight into what it’s like attending MIT. Some of the first students Singhal mentored through the program are now also participants as well.

    “The further you get in your educational career, the more you realize how much assistance you got along the way to get where you are,” says Singhal. “That happens at every stage.”

    Looking toward the future, Singhal wants to continue to pursue research with a sustainability impact. She also wants to continue mentoring in some capacity but isn’t in a rush to figure out exactly what that will look like.

    “Grad school doesn’t mean I have to do one thing. I can stay open to all the possibilities of what comes next.”  More