Artificial intelligence research is constantly developing new concepts that have the potential to benefit society and industry; However, sometimes these benefits are not fully realized due to lack of engineering equipment. To help bridge this gap, graduate students in the MIT Department of Electrical Engineering and Computer Science’s 6-A Master of Engineering (MEng) thesis program work with some of the world’s most innovative companies and contribute cutting-edge projects. cooperate on. to and to complete their MEng thesis.
During part of the past year, four 6-A MEng students completed an internship with IBM Research’s advanced prototyping team through the MIT-IBM Watson AI Lab on AI projects, often designed to solve a real-world issue. Develops web applications for or commercial use cases. Students here work with AI engineers, user experience engineers, full-stack researchers, and generalists to accommodate project requests and receive thesis advice, says Lee Marty, IBM Research staff member and 6-A manager. Students’ projects range from generating synthetic data to using computer vision to allow privacy-sensitive data analysis to identify tasks in video that require monitoring human safety and tracking construction progress at the construction site. allows.
“I appreciate all the expertise and feedback from the team,” says 6-A graduate Violetta Zusiga ’21, who participated in the program. “I think working in industry gives the lens of making sure that the project needs and [provides the opportunity] To do ground research and make sure it is helpful for any future use case. ,
Jusiega’s research intersects the fields of computer vision and design to focus on data visualization and user interfaces for the medical field. Working with IBM, she created an application programming interface (API) that lets physicians interact with an AI model of a medical treatment strategy, which was deployed in the cloud. Its interface provided a medical decision tree, as well as some prescribed treatment plans. After receiving feedback on its design from physicians at a local hospital, Zusiga developed iterations of the API and how the results were displayed, visually, so that it was user-friendly and understandable to physicians who typically do not code. . She adds that, “these tools are often not acquired in the field because they lack some of these API principles which become more important in an industry where everything already happens at a very fast pace.” , so there is little time to incorporate new technology.” But this project may eventually allow for industry deployment. “I think this application has a bunch of potential, whether it’s taken up by clinicians or whether it’s just used in research. It’s very promising and very exciting to see how technology can help us modify What can help, or I can improve, is make the health care sector even more custom-tailored towards patients and provide them with the best care possible,” she says.
Another 6-A graduate student, Spencer Compton, was looking at helping professionals make more informed decisions for use in settings including health care, but he was tackling it from a causal perspective. When given a set of related variables, Compton was investigating whether there was a way to determine not only the correlation, but the cause-and-effect relationship between them (the direction of the interaction) from the data alone. For this, he and his colleagues from IBM Research and Purdue University turned to a field of mathematics called information theory. With the goal of designing an algorithm for learning complex networks of causal relationships, Compton used ideas related to entropy, the randomness in a system, to help determine whether a causal relationship exists and How variables can interact. “When deciding on an explanation, people often default to Occam’s razor” Compton says. “We are inclined to believe the simpler explanation than the more complex one.” In many cases, he says, it looked like it was performing well. For example, they were able to consider variables such as lung cancer, pollution and X-ray findings. He was pleased that his research allowed him to help build a framework of “entropic causal inference” that could satisfactorily aid in safe and smart decisions in the future. “The math is actually surprisingly deep, interesting, and complex,” says Compton. “We’re basically asking, ‘When is the simplest explanation correct?’ But as a question of mathematics.”
Determining relationships within data can sometimes require a large amount of time to decipher patterns, but this may not be available for data containing sensitive information. For her master’s work, Ivy Huang worked with IBM Research to generate synthetic tabular data using a natural language processing tool called the Transformer Model, which can learn from past values and predict future values. Trained on real data, the model can produce new data with similar patterns, properties and relationships without the restrictions such as confidentiality, availability and accessibility that can come with real data in financial transactions and electronic medical records. In addition, it created an API and deployed the model to an IBM cluster, which allowed users to extend access to the model and capabilities without compromising on the original data.
Working with the Advanced Prototyping team, MEng candidate Brandon Perez also considered how to collect and calibrate data with no restrictions, but in his case a computer focused on an action recognition model, to identify construction site incidents. Had to use the vision framework. The team based their work on the Moments in Time dataset, which contains more than a million three-second video clips with about 300 attached classification labels, and performed well during AI training. However, the group needed more construction-based video data. For this he used YouTube-8M. Perez created a framework for testing and fine-tuning existing object detection models and action recognition models that could plug into an automated spatial and temporal localization tool – how they would identify and label particular actions in a video timeline. “I was satisfied that I was able to figure out what made me curious, and I’m grateful for the autonomy I was given with this project,” Perez says. “I felt like I was always supported, and my mentors were a huge support to the project.”
“The kind of collaboration we’ve seen between our MEng students and IBM researchers is about the 6-A Meng thesis program at MIT,” says Tomas Palacios, professor of electrical engineering and faculty director at MIT 6-A. MEng Thesis Programme. “For more than 100 years, 6-A has been connecting MIT students with industry to solve some of the world’s most pressing problems together.”
