http://hdl.handle.net/1957/8305 Search the Channel search http://ir.library.oregonstate.edu/jspui/simple-search http://hdl.handle.net/1957/13549 Title: Asymmetric Tensor Visualization with Glyph and Hyperstreamline Placement on 2D Manifolds<br/><br/>Authors: Palke, Darrel; Chen, Guoning; Lin, Zhongzang; Yeh, Harry; Laramee, Robert; Zhang, Eugene<br/><br/>Abstract: Asymmetric tensor fields present new challenges for visualization techniques such as hyperstreamline placement and glyph packing. This is because the physical behaviors of the tensors are fundamentally different inside real domains where eigenvalues arereal and complex domains where eigenvalues are complex. We present a hybrid visualization approach in which hyperstreamlines are used to illustrate the tensors in the real domains while glyphs are employed for complex domains. This enables an effective visualization of the flow patterns everywhere and also provides a more intuitive illustration of elliptical flow patterns in the complex domains. The choice of the types of representation for different types of domains is motivated by the physical interpretation of asymmetric tensors in the context of fluid mechanics, i.e., when the tensor field is the velocity gradient tensor. In addition, we encode the tensor magnitude to the size of the glyphs and density of hyperstreamlines. We demonstrate the effectiveness of our visualization techniques with real-worldengine simulation data. Tue, 08 Dec 2009 14:24:44 GMT http://hdl.handle.net/1957/13225 Title: End-User Feature Engineering in the Presence of Class Imbalance<br/><br/>Authors: Oberst, Ian; Moore, Travis; Wong, Weng-Keen; Kulesza, Todd; Stumpf, Simone; Riche, Yann; Burnett, Margaret<br/><br/>Abstract: Intelligent user interfaces, such as recommender systems and email classifiers, use machine learning algorithms to customize their behavior to the preferences of an end user. Although these learning systems are somewhat reliable, they are not perfectly accurate. Traditionally, end users who need to correct these learning systems can only provide more labeled training data. In this paper, we focus on incorporating new features suggested by the end user into machine learning systems. To investigate the effects of user-generated features on accuracy we developed an auto- coding application that enables end users to assist a machine-learned program in coding a transcript by adding custom features. Our results show that adding user-generated features to the machine learning algorithm can result in modest improvements to its F1 score. Further improvements are possible if the algorithm accounts for class imbalance in the training data and deals with low-quality user-generated features that add noise to the learning algorithm. We show that addressing class imbalance improves performance to an extent but improving the quality of features brings about the most beneficial change. Finally, we discuss changes to the user interface that can help end users avoid the creation of low-quality features. Mon, 02 Nov 2009 22:44:04 GMT http://hdl.handle.net/1957/12770 Title: A Strategy-Centric Approach to the Design of End-User Debugging Tools<br/><br/>Authors: Grigoreanu, Valentina I.; Burnett, Margaret; Robertson, George<br/><br/>Abstract: End-user programmers’ code is notoriously buggy. This problem is amplified by the increasing complexity of end users’ programs. To help end users catch errors early and reliably, we employ a novel approach for the design of end-user debugging tools: a focus on supporting end users’ effective debugging strategies. This paper has two core contributions. We first demonstrate the potential of a strategy-centric approach to tool design by presenting StratCel, a strategy-based tool for Excel. Second, we show the benefits of this design approach: participants using StratCel found twice as many bugs as participants using standard Excel, they fixed four times as many bugs, and all this in only a small fraction of the time. Furthermore, this strategy-based approach helped the participants who needed it the most: boosting novices’ debugging performance near experienced participants’ improved levels. Finally, we reveal several opportunities for future research about strategy-based debugging tools. Thu, 24 Sep 2009 19:30:10 GMT http://hdl.handle.net/1957/12706 Title: End-User Debugging of Machine-Learned Programs: Toward Principles for Baring the Logic<br/><br/>Authors: Kulesza, Todd; Stumpf, Simone; Riche, Yann; Burnett, Margaret; Wong, Weng-Keen; Oberst, Ian; Moor, Travis; McIntosh, Kevin; Bice, Forrest<br/><br/>Abstract: Many applications include machine learning algorithms intended to learn “programs” (rules of behavior) from an end user’s actions. When these learned programs are wrong, their users receive little explanation as to why, and even less freedom of expression to help the machine learn from its mistakes. In this paper, we develop and explore a set of candidate principles for providing salient debugging information to end users who would like to correct these programs. We informed the candidate principles through a formative study, built a prototype that instantiates them, and conducted a user study of the prototype to collect empirical evidence to inform future variants. Our results suggest the value of exposing the machine’s reasoning process, supporting a flexible debugging vocabulary, and illustrating the effects of user changes to the learned program’s logic. Mon, 21 Sep 2009 13:39:33 GMT