Reference Information:
Title: Real-time Interaction with Supervised Learning
Author: Rebecca Fiebrink
Venue: CHI 2010
Summary:
The author begins by explaining the process that defines machine learning. She then explains that her work is specifically on improving a musical machine learning program. She discusses how programs on machine learning come from engineers and do not receive as much attention from musicians during creation. She therefore studied musicians using this program and came up with ways to improve it. Three ideas she came up with were interactive editing of training data, supporting use by non-experts, and real user priorities. In the future she wishes to do a more formal study of her improvements on the system.
Discussion:
I enjoyed reading about the work that the author is doing. One aspect of this work I like is that it is a creative use of engineering, much like dance.draw. Another aspect is the study of interfaces for people who are not engineers. I think it can be difficult to make programs for those who are not frequent users of technology. By studying those who are not good with technology we can find better ways to create interfaces for them.
Definitely the most interesting part of this is how she was able to incorporate users that were not explicitly computer scientists and make it user-friendly for them. I think that there is room to improve on this to make it even more fluid, but I have not used the program so it would be hard to say specifically how.
ReplyDeleteThis was my thought exactly. I like how she attempts to incorporate very abstract areas of computer science (machine learning) with other concrete and creative areas.
ReplyDelete