Graduate Thesis Or Dissertation
 

Incorporating User Feedback into Machine Learning Systems

Pubblico Deposited

Contenuto scaricabile

Scarica il pdf
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/9019s7533

Descriptions

Attribute NameValues
Creator
Abstract
  • Although machine learning systems are often effective in real-world applications, there are situations in which they can be even better when provided with some degree of end user feedback. This is especially true when the machine learning system needs to customize itself to the end user's preferences, such as in a recommender system, an email classifier or an anomaly detector. This thesis explores two directions in incorporating end user feedback to machine learning systems. First, I introduce an algorithm that incorporates feature feedback in a semi-supervised text classification setting. Feature feedback goes beyond instance-label feedback by allowing end users to indicate which feature-value combinations are predictive of the class label. In order to incorporate feature feedback in a semi-supervised setting, I develop a Locally Weighted Logistic Regression algorithm that uses a similarity metric combining information from the user's feature feedback and information based on label diffusion on the unlabeled data. Second, I explore the use of instance-level feedback to anomaly detection algorithms. Anomaly detectors commonly return a list of the top outliers in the data. Although these outliers are statistically unusual, some are uninteresting to a user as the internal statistical model may not necessarily be aligned with the user's semantic notion of an anomaly. I present an algorithm that can increase the number of true anomalies presented to the user if a limited amount of instances are labeled as anomalous or nominal.
License
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Dichiarazione dei diritti
Publisher
Peer Reviewed
Language
Replaces

Le relazioni

Parents:

This work has no parents.

In Collection:

Elementi

Thumbnail Titolo Data caricata Visibilità Azioni
file details DasShubhomoy2017.pdf 2017-11-21 Pubblico Scaricare