ICCV Daily 2019 - Tuesday

Bachelor’s in information technology and engineering in Latina. First of all, it was my parents pushing me for the information technology branch because I have a propension for doing things with the computer. There, I met my professor of machine learning, Barbara Caputo, whom I really want to mention because I think she is great. She is actually the one who introduced me to machine learning. She made me like it a lot. After that I started a course in Rome in artificial intelligence and robotics but, at the same time, I started looking around. There was a course in Cambridge that was a new experimental course. It was the second year it was running. It's called “Master of Philosophy in Machine Learning, Speech, and Language Technology”. I looked a bit at the program, and I thought it would be great if I could join the program. There was a long submission process and a long application process. I finally ended up being in this group, and I worked there for one year. I studied for months, and then I worked in research. That's where I met my other professor, Mark Gales. He is doing some work on the interpretability of neural networks for speech recognition and language processing. I did my Master’s thesis project with him, and I found it quite interesting. I wanted to look into other applications for this. In between the Master’s in Rome and a Master’s in Cambridge, I went to Switzerland, 16 Women in C.Vision DA I L Y always through Barbara. I went to visit for an internship of two months at the lab of Henning Müller. I thought it was a great place, and we did great work. I think we published something that was accepted. We studied a totally different case that was an analysis of surface electromyography signals. Of all these things that you learned from Barbara, that she pushed you in a certain direction, is there anything that struck you in particular? Yes, but it's not her sentence. it's a Tom Mitchell sentence. Everything is copied in this world! [laughs] I think that it is a definition of machine learning. It's something like: "A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E." The essence is that you can develop an algorithm that learns with experience. So the more data you give to the algorithm, the more it learns. Aren’t you the same? The more you have experience, the more you learn, the more you know? Well, I think yes, kind of everybody learns from experience.