Machine learning a source of inspiration
A Binghamton undergraduate says he sees understanding data as the future of helping others.
Adiel Felsen, a junior double majoring in computer science and studies machine learning.
鈥淢achines make mistakes but they make mistakes less often, especially if they鈥檙e trained properly,鈥 Felsen said. 鈥淎nd there can be a lot of dangers in that if you start fully relying on only machines, but I really think machine learning is a great way forward for science in general.鈥
He has always had an interest in doing research, which played a part in his attraction to Binghamton and the First-Year Research Immersion program.
鈥淚 was interested before I went into the program, because the program is part of what attracted me to Binghamton,鈥 Felsen said. 鈥淏ut it definitely allowed me to learn about topics that I wouldn鈥檛 otherwise learn until maybe senior year of school.鈥
FRI helped Felsen develop an understanding of data and machine learning. He categorizes data as anything from random images to letters and numbers. As a part of FRI, he worked on computer vision, a field of study in which researchers teach computers how to complete simple tasks such as determining whether an image shows a cat or dog.
During his sophomore year, Felsen returned to FRI as a teaching assistant. In that role, he not only assisted other students in understanding the basics of how to conduct research, he also solidified his own love for research.
Now working under Associate Professor of Computer Science Kenneth Chiu and Assistant Professor of Biomedical Engineering Frank Lu, Felsen conducts research on biomedical imaging. He and his lab colleagues segment nuclei from cancer cells for SRS imaging.
SRS imaging, or stimulated Raman scattering microscopy, provides a detailed view of a cell. The method does not require the cell to be stained or dyed. Although it鈥檚 very fast, SRS imaging does not capture good contrast of DNA, and in biomedical imaging researchers need to be able to detect cell nuclei. Through machine learning algorithms, Felsen and his colleagues can improve the images鈥 contrast and see the nuclei.
One of Felsen鈥檚 mentors describes him as dynamic. 鈥淗e has persistence and is always exploring new ideas,鈥 Chiu said. 鈥淭hose are the key things for research.鈥
Felsen will continue to develop his understanding of data and its many uses this summer as an intern at Mars Inc. He鈥檒l work on projects dealing with data science for the chocolate company.
Felsen does not allow himself to be frustrated by minor setbacks because of what he sees as the impact of his work.
鈥淥ur research has the potential to help speed up the process of brain surgery, and it likely has a long way to go before it is actually implemented in the operating room,鈥 Felsen said. 鈥淗owever, our research demonstrates that machine learning algorithms can improve the contrast of SRS images to look similar to stained cells. Hopefully, other researchers will see this promise and will be inspired to pursue similar work.鈥
Editor鈥檚 note: This article appeared in .