Reference Information:
Title: Lowering the Barrier to Applying Machine Learning
Author: Kayur Patel
Venue: CHI 2010
Summary:
The author begins by describing some of the problems with learning based development software. One such problem is the difficulty that researchers have understanding the relationship between the accuracy and different parts of the machine learning system. The author argues that problems such as this result from the lack of flexibility that current machine learning software provides. As a result the author has begun the creation of Gestalt, a general environment supporting the application of learning techniques. Gestalt will offer greater flexibility and visualization in order to make machine learning projects easier.
Discussion:
While I do find machine learning interesting, I think that this paper should have been written after she had achieved more results from her Gestalt environment. A great deal of this paper is about what she hopes Gestalt may one day achieve instead of what it actually has achieved. I did enjoy her explanation on how in some situations accuracy should be lowered to heighten performance time. In future work I hope to see a more established system with more concrete results.
I respectfully disagree and I am going to give the author the benefit of the doubt. I would say that a lot of research has been proven only on test subjects and such, therefore its true potential is still unknown.
ReplyDeleteI agree with this to a point, I do agree that it is strange that the author is only writing about what they want the program to do and not what it does, however you have to remember this is a dissertation introduction. I also agree its an interesting idea, and I will also agree with Angel and say that the author did do a lot of research and needs to qualify the study.
ReplyDeleteI wonder in what sorts of applications the author prefers speed to accuracy. Let us hope she emphasizes accuracy on the compiler side.
ReplyDeleteI had the same frustration with the paper in that it seemed to have a good statement of intent, but didn't offer much substantive comment with regards to results. It did leave me curious as to how this is all going to play out and I'm interested in hearing a summary of the results
ReplyDelete