Tag

research

Using GenAI to design floor plans and buildings

By | Data, Data sets, Educational, Feature, General Interest, HPC, News, Research, Systems and Services

There is a lot to consider when designing places where humans live and work. How will the space be used? Who’s using the space? What are budget considerations? It is painstaking and time consuming to develop all of those details into something usable. 

What if Generative AI (GenAI) could help? We already know that it can be used to create text, music, and images. Did you know that it can also create building designs and floor plans? 

Dr. Matias del Campo, associate professor of architecture in the Taubman College for Architecture and Urban Planning, has been working to make architectural generative models more robust. He aims to expand on the patterns, structures, and features from the available input data to create architectural works. Himself a registered architect, designer, and educator, del Campo conducts research on advanced design methods in architecture using artificial intelligence techniques.

He leverages something called neural networks for two projects: 

  • Common House: A project that focuses on floor plan analysis and generation.
  • Model Mine: A large-scale, 3D model housing database for architecture design using Graph Convolutional Neural Networks and 3D Generative Adversarial Networks.

This is an example from the annotated data created from the Common House research project. The main obstacle that has emerged in creating more real-life plans is the lack of databases that are tailored for these architecture applications. The Common House project aims at creating a large-scale dataset for plans with semantic information. Precisely, our data creation pipeline consists of annotating different components of a floor plan, for e.g., Dining Room, Kitchen, Bed Room, etc.

 

Four quadrants showing 9 models each of chairs, laptops, benches, and airplanes

A large scale 3D model housing database for Architecture design using Graph Convolutional Neural Networks and 3D Generative Adversarial Networks.

What exactly are neural networks? The name itself takes inspiration from the human brain and the way that biological neurons signal to one another. In the GenAI world, neural networks are a subset of machine learning and are at the heart of deep learning algorithms. This image of AI hierarchy may be helpful in understanding how they are connected.

Dr. del Campo’s research uses GenAI for every step of the design process including 2D models for things like floors and exteriors, and 3D models for shapes of the rooms, buildings, and volume of the room. The analysis informs design decisions. 

DEI considerations

del Campo notes that there are some DEI implications for the tools he’s developing. “One of the observations that brought us to develop the ‘Common House’ (Plangenerator) project is that the existing apartment and house plan datasets are heavily biased towards European and U.S. housing. They do not contain plans from other regions of the world; thus, most cultures are underrepresented.” 

To counterbalance that, del Campo and his team made a global data collection effort, collecting plans and having them labeled by local architects and architecture students. “This not only ensured a more diverse dataset but also increased the quality of the semantic information in the dataset.”

How technology supports del Campo’s work

A number of services from Information Technology & Services are used in these projects, including: Google at U-M collaboration tools, GenAI, Amazon Web Services at U-M (AWS), and GitHub at U-M

Also from ITS, the Advanced Research Computing (ARC) team provides support to del Campo’s work. 

“We requested allocations from the U-M Research Computing Package for high-performance computing (HPC) services in order to train two models. One focuses on the ‘Common House’ plan generator, and the other focuses on the ‘Model Mine’ dataset to create 3D models based,” said del Campo. 

Additionally, they used HPC allocations from the UMRPC in the creation of a large-scale artwork called MOSAIK which consists of over 20,000 AI-generated images, organized in a color gradient. 

A large scale 3D model housing database for Architecture design using Graph Convolutional Neural Networks and 3D Generative Adversarial Networks.

“We used HPC to run the algorithm that organized the images. Due to the necessary high resolution of the image, this was only possible using HPC.”

“Dr. del Campo’s work is really novel, and it is different from the type of research that is usually processed on Great Lakes. I am impressed by the creative ways Dr. del Campo is applying ITS resources in a way that we did not think was possible,” said Brock Palen, director of the ITS Advanced Research Computing. 

Related: Learn about The Architecture + Artificial Intelligence Laboratory (AR2IL)

PFAS research in the Michigan mother-infant pairs study, supported by ITS, SPH, MM, AGC

By | News

Three mothers holding their infants. Everyone is sitting on a couch..PFAS (per- and polyfluoroalkyl substances) are a class of chemicals that have been around since the 1940s and became more broadly used in the post-war 1960s era. PFAS are in our homes, offices, water, and even our food and blood. PFAS break down slowly and are difficult to process, both in the environment and our bodies. 

Scientific studies have shown that exposure to some PFAS in the environment may be linked to harmful health effects in humans and animals. Because there are thousands of PFAS chemicals found in many different consumer, commercial, and industrial products, it is challenging to study and assess the human health and environmental risks. 

Fortunately, some of the most persistent PFAS are being phased out. The EPA has been working on drinking water protections, scientists are working on ways to break down and eliminate PFAS, and PFAS are being addressed at a national level

A team of University of Michigan researchers from the School of Public Health DoGoodS-Pi Environmental Epigenetics Lab and Michigan Medicine are working to understand how behaviors and environments during pregnancy can cause changes to the way genes work in offspring. This emerging field is known as toxicoepigenetics. 

Jackie Goodrich, Ph.D., research associate professor at the U-M School of Public Health, led the team. “PFAS may impact the development of something we all have called the epigenome. The epigenome is a set of modifications on top of our DNA that controls normal development and function. Environmental exposures like PFAS can alter how the epigenome forms, and this impacts development and health. Our study expands on current knowledge about PFAS and the epigenome by focusing on a type of epigenetic mark that is not usually measured.”

Vasantha Padmanabhan, Ph.D., M.S., professor emerita (in service), Department of Pediatrics, Michigan Medicine, built the Michigan Mother-Infant Pairs study over the past decade with an emphasis on identifying harmful exposures during pregnancy that impact women and their newborns. “I am so grateful to those who engaged in this study. PFAS are complex, and mothers’ and infants’ involvement helped us work toward a solution that impacts us all. I want to acknowledge the contributions of the U-M Department of Obstetrics and Gynecology, Michigan Institute for Clinical & Health Research (MICHR), and the Von Voigtlander Women’s Hospital that made this study possible.” 

Rebekah Petroff, Ph.D., a research fellow with Environmental Health Sciences, led the computation portion of the research. She said that using Turbo for storing the raw data and Great Lakes for high-performance computing (HPC) enabled a much faster analysis that was needed for the study with so much data to analyze. 

Turbo and Great Lakes are services provided by Advanced Research Computing, a division of Information and Technology Services (ITS). ARC facilitates powerful approaches to complex research challenges in fields ranging from physics to linguistics, and from engineering to medicine.

Petroff said, “This analysis would have taken over a month straight of computing time on a regular desktop computer. The first job we submitted to Great Lakes ran so fast—I had results the next morning! Great Lakes made this research possible, and I believe that our study results can be broadly impactful to public health and toxicoepigenetics going forward.”

Support for using this complex technology also came from Dan Barker, a UNIX systems admin with the U-M School of Public Health Biostatistics Department. Barker assisted with the code needed to use Great Lakes. “We started with a test run of a few hundred pairs of genomes. Once we were successful with that, we ran the entire nearly 750,000 epigenetic marks across 141 people and seven different PFAS.”

Barker also helped design and submit array jobs which are a series of identical, or near identical, tasks that are run multiple times. This is a common technique used by researchers when leveraging HPC. Array jobs allow for essential analytical comparisons among the test results. Petroff said, “In our study, we used an array job to split up our computations so that they ran much more efficiently!”

The U-M Advanced Genomics Core (AGC) performed the epigenetic assays, a kind of laboratory technique which measures marks on your DNA, for this project. AGC is part of the campus-wide laboratories that develop and provide state-of-the-art scientific resources to enable biomedical research known as Biomedical Research Core Facilities (BRCF). Other BRCF cores also worked on this project, including the Epigenomics Core and the Bioinformatics Core.

Genotyping is similar to reading a few words scattered on a page. This process gives researchers small packets of data to compare. Genotyping looks for information at a specific place in the DNA where we know important data will be. This project used a type of genotyping called microarrays (also known as “arrays”) and help researchers understand how regulation of DNA—including methylation and hydroxymethylation measured in this study—are impacted by exposures like PFAS.  

Brock Palen, ARC director, said, “This research is of human interest and impacts all of us. I’m pleased that ARC assisted their research with staff expertise, equipment, and no-cost allocations from the U-M Research Computing Package.”

Petroff said that follow up studies are needed to better understand if the results are universal or specific to this cohort of infants and parents. If the results hold steady, then a significant discovery has been made that will lead to more comprehensive PFAS mitigation solutions. “Although steps are being taken to mitigate PFAS, exposure is still prevalent, and a deeper understanding of how it impacts humans is needed,” said Dana Dolinoy, Ph.D., chair, NSF International Department Chair of Environmental Health Sciences and epigenetics expert.

Read the full article: Mediation effects of DNA methylation and hydroxymethylation on birth outcomes after prenatal per- and polyfluoroalkyl substances (PFAS) exposure in the Michigan mother–infant pairs cohort.

Funding was provided by grants from the National Institutes of Health, the U.S. Environmental Protection Agency, and the National Institute of Environmental Health Sciences Children’s Health Exposure Analysis Resource program.

Understanding the strongest electromagnetic fields in the universe

By | Data, Great Lakes, HPC, Research, Uncategorized

Alec Thomas is part of the team from the U-M College of Engineering Gérard Mourou Center for Ultrafast Optical Science that is building the most powerful laser in the U.S.

Dubbed “ZEUS,” the laser will be 3-petawatts of power. That’s a ‘3’ with 15 zeros. All the power generated in the entire world is 10-terawatts, or 1000 times less than the ZEUS laser. 

The team’s goal is to use the laser to explore how matter behaves in the most extreme electric and magnetic fields in the universe, and also to generate new sources of radiation beams, which may lead to developments in medicine, materials science, and national security. 

A simulation of a plasma wake.

This simulation shows a plasma wake behind a laser pulse. The plasma behaves like water waves generated behind a boat. In this image, the “waves” are extremely hot plasma matter, and the “boat” is a short burst of powerful laser light. (Image courtesy of Daniel Seipt.)

“In the strong electric fields of a petawatt laser, matter becomes ripped apart into a `plasma,’ which is what the sun is made of. This work involves very complex and nonlinear physical interactions between matter particles and light. We create six-dimensional models of particles to simulate how they might behave in a plasma in the presence of these laser fields to learn how to harness it for new technologies. This requires a lot of compute power,” Thomas said. 

That compute power comes from the Great Lakes HPC cluster, the university’s fastest high-performance computing cluster. The team created equations to solve a field of motion for each six-dimensional particle. The equations run on Great Lakes and help Thomas and his team to learn how the particle might behave within a cell. Once the field of motion is understood, solutions can be developed. 

“On the computing side, this is a very complex physical interaction. Great Lakes is designed to handle this type of work,” said Brock Palen, director of Advanced Research Computing, a division of Information and Technology Services. 

Thomas has signed up for allocations on the Great Lakes HPC cluster and Data Den storage. “I just signed up for the no-cost allocations offered by the U-M Research Computing Package. I am planning to use those allocations to explore ideas and concepts in preparation for submitting grant proposals.”

Learn more and sign up for the no-cost U-M Research Computing Package (UMRCP).

Prof. Thomas’ work is funded by a grant from the National Science Foundation.

MICDE to provide data analysis and dissemination support for $18 million tobacco research center

By | General Interest, Happenings, News, Research

The University of Michigan School of Public Health will house a new, multi-institutional center focusing on modeling and predicting the impact of tobacco regulation, funded with an $18 million federal grant from the National Institutes of Health and the Food and Drug Administration.

The Center for the Assessment of the Public Health Impact of Tobacco Regulations will be part of the NIH and FDA’s Tobacco Centers of Regulatory Science, the centerpiece of an ongoing partnership formed in 2013 to generate critical research that informs the regulation of tobacco products.

The Michigan Institute for Computational Discovery and Engineering (MICDE) will support the center’s Data Analysis and Dissemination core by collecting national and regional survey data, conducting analysis of the use of tobacco products including vaping and e-cigarettes, and disseminate the resulting tobacco modeling parameters to other research centers and the Food and Drug Administration.

The center is led by MICDE affiliated faculty member Rafael Meza, associate professor of Epidemiology, and David Levy, professor of Oncology at Georgetown University.

For more on the center, see the press release from the U-M School of Public Health: https://sph.umich.edu/news/2018posts/tcors-091718.html

Video available from MIDAS Research Forum

By | General Interest, Happenings, News, Research

Video is now available from the MIDAS Research Forum held Dec. 1 in the Michigan League at http://myumi.ch/6vA3V

The forum featured U-M students and faculty showcasing their data science research; a workshop on how to work with industry; presentations from student groups; and a summary of the data science consulting and infrastructure services available to the U-M research community.

NOTE: The keynote presentation from Christopher Rozell of the Georgia Institute of Technology will be available in the near future.

2017 U-M Data Science Research Forum

By |

Forum Highlights

  • Oral and poster presentations on
    • Theoretical foundations of data science
    • Data science methodology
    • Data science applications in any research domain
    • Social impact of data science research
  • Networking Reception

All presentations will come from submissions in response to our call for abstracts
Abstract Submission Deadline: October 23, 2017
We welcome submission from all U-M data science researchers (faculty, staff, trainees)

Please register for this event.  Please also see the call for abstracts for instruction, and submit through the Abstract Submission Form.

Preliminary Schedule