Here at the University of Connecticut, the College of Engineering’s annual poster competition displayed over 100 posters on the wide array of research engineering graduate students have been working on. With projects ranging from electronic noses to logic gates operated via sound, there was something for everyone.
Hosted in the Student Union Ballroom on March 4, students from a range of disciplines had five minutes to present their findings to a group of judges, with three minutes to present their research and two minutes to answer any pressing questions.
Aida Ghiaei, graduate outreach manager for the College of Engineering, said the idea behind the competition is based on the Three Minute Thesis competition held at research institutions around the world.
“They should be able to explain it as [if you’re] sitting at… dinner and telling your aunt what you are doing in three minutes,” Ghiaei said.
One of the graduate students presenting was Khoa Nguyen, from the School of Mechanical, Aerospace and Manufacturing Engineering (SoMAM). Nguyen’s research used metamaterials to create transistors that can operate via sound. Metamaterials are engineered materials whose properties come from the design of its internal structure, rather than the chemical composition of its base substances.
Since silicon transistors only work in special environments, Nguyen tried to see if he could use metamaterials to create transistors that worked outside of these environments. He used them to create “building blocks” that he arranged into four sets, while using a magnet to control the direction the open face of these blocks face. These magnets are controlled by a sound at a set frequency.
Based on how the magnets are arranged — either connecting two sets vertically or horizontally — Nguyen is able to create logic gates whose signal — either zero or one — depends on the direction of these blocks.
Andrew Shortridge, a graduate student in the materials science department, focused his research on a laser — the strongest laser in New England, according to Shortridge — to detect the structural effects of high pressure and temperatures on materials computational framework to model the results. He described the pockmarks on the steel square he held in his hand as “laser shocks” from the experiment. Shortridge said he confirmed that their results also “lined up with existing literature,” implying that their results are valid with other experiments of the same nature.
With a poster titled “One vector to rule them all,” computer science PhD candidate Kaiyang Li explained how linear algebra can reduce the number of parameters a large language model (LLM) uses from over 7 billion to under a million. Parameters are the settings that control and optimize an LLM’s output and behavior, according to IBM.
By using Uni-LoRA (Low-Rank Adaptation), a method to fine-tune the number of parameters an LLM uses, you can flatten one vector into a matrix and, by transforming it using a projection matrix, you can improve the parameter efficiency, according to Li. Even with fewer parameters, the Uni-LoRA model achieved similar performance to larger LLM at a reduced computational cost. You can find the source code for this framework online.
Delowar Hossain is another researcher whose project uses an artificial intelligence model, this time to train an electronic nose to detect chemical scents. Packed with over 70 sensors that measure electrical resistance, Hossain provided the nose training data to let it formulate its own idea of over a dozen chemical scents, including chloroform, hexane and pentanol.
The nose was able to use spider graphs to predict each scent and had 95-100% accuracy in predicting the correct scent, albeit in dry air. Hossain hopes to improve the electronic nose’s ability to detect scents in the presence of moisture.
After an hour of perusing posters, all the judges left to deliberate on the first round of winners across each department.
The department winners were Atta Seck and Cao Sang Troung for biomedical engineering, Maryam Mohebi for chemical and biomolecular engineering, Adanma Akoma for materials science and engineering, Ana Carolina Vieira Rocha and Dahye Kim for the civil engineering program in the School of Civil and Environmental Engineering, Samuel Rothfarb for the environmental engineering program in the same school, Giovanni Romano for the School of Computing, Blake Gaines for electrical and computer engineering and Melanie Keogh for SoMAM.
Ayah Abdallah and Mohebi took home the People Choice awards, which people voted on by scanning the QR codes next to their favorite poster. Each of the winners listed above won $50 in prize money.
After a second round of deliberation among the judges, the top three winners of the competition were announced: Rothfarb, Akoma and Carolina Vieira Rocha took home $1,000, $500 and $250 respectively.
Ghiaei says the aim of these poster competitions is to “expose the research we are doing to [the] industry, hopefully they [students] can make connections … [it] also levels up the visibility of College of Engineering research.”
Her favorite part about these events is seeing “the energy that the students bring in and the excitement, the nervousness and happiness,” Ghiaei said she also appreciates the improvement she has seen from some graduate students who return year after year.
“The students, some of them, come year after year and I see how they improve, I think that’s really fun to watch,” Ghiaei said.
When asked about what makes a good poster presentation, Ghiaei said the students need to show enthusiasm, be aware of the language they use and be open to any questions.
“Research is about the questions…The excitement that the person shows and the less jargon they use and [how] they really grab your attention is important,” Ghiaei said.