Some of the avi files require the TechSmith Screen Capture Codec. Play these (and RealMedia files) at exactly 100% to retain lossless quality.
Mark Derthick and John Zimmerman. The Perspectives Browser: Exploratory Data Analysis for Everyone. Submitted to the IEEE Information Visualization Symposium (InfoVis'05). Accompanying 3 minute video (14MB avi).
Finding patterns in databases of political contributions, pollution and environmental data, or hospital and school performance would surely interest many citizens. The Perspectives Browser is intended to support this kind of exploration for users with little or no training in statistics or programming. Given an "advanced search" type query, it visualizes dependencies on the query of up to 30 variables. In preliminary studies, participants found interesting three-variable dependencies in an art collection.
Mark Derthick. Incremental Visual Queries. Submitted to ACM Transactions on Information Systems. Accompanying 4 minute video: 2MB RealMedia or 3MB avi
The semantics of traditional database query languages like SQL is confined to individual queries. In interactive systems, a more complete account also models the relationship between successive queries. This paper uses the metaphor of navigation to begin with one query and instantiate a related query in one step. Structurally complicated SQL queries can be built up incrementally with multiple navigation operations. Each component query represents an intermediate result, which reinforces the compositional semantics.
Mark Derthick and Stephen F. Smith. An Interactive 3D Visualization for Requirements Analysis.. To appear in Journal of Scheduling. Accompanying 4 minute video: 8MB avi file, 6MB Quicktime, or 4MB RealMedia
AI scheduling systems assign a given set of resources to activities over time to satisfy a given set of demands. However analysts often must also negotiate requirements and system capabilities. Current schedulers poorly support this higher level task, because lateness in one particular schedule does not necessarily lend insight into the tradeoffs in constraint space. The most important variables for each demand are the total weight of the cargo and the time window between when the cargo becomes available at the port of origin and when it is due at the destination port. Therefore a 3D visualization can summarize the constraint space. The analyst can see constraint violations at a glance, and do "what if" analyses in seconds.
Mark Derthick, Michael G. Christel , Alexander G. Hauptmann, Howard D. Wactlar. Constant Density Displays Using Diversity Sampling. In Proceedings of the IEEE Information Visualization Symposium (InfoVis'03), Seattle, WA, pp. 137-144. Accompanying 3 minute video: 9MB avi file or 3MB RealMedia
For a web-style query on "jaguar", presentation by ranking alone might list 100s of car-related sites before any animal-related sites. Making a more diverse set of results more salient gives a better overview for browsing tasks. This kind of sampling requires more computation, because the score for one sample depends on all the other samples. A greedy algorithm is presented that is fast enough to support direct manipulation from a dataset of 350,000 video keyframes.
Mark Derthick, Michael G. Christel, Alexander G. Hauptmann, Dorbin Ng, Scott Stevens, Howard D. Wactlar. A Cityscape Visualization of Video Perspectives. Poster presented at the National Academy of Sciences Arthur M. Sackler Colloquium on Mapping Knowledge Domains, Irvine, CA, 2003.
This poster suggests an approach to visualizing objects along 8-20 meaningful dimensions simultaneously by combining the idea of parallel coordinates with cityscapes. Buildings represent parallel axes on which images representing the objects are displayed. Spotlights correspond to the line segments of parallel coordinates. Three navigable dimensions provide more space for data display, interaction, and links between more pairs of dimensions.
Mark Derthick. Interfaces for Palmtop Image Search. In Proceedings of the Joint ACM/IEEE Conference on Digital Libraries, Portland, OR, 2002, pp. 340-341.
Current image retrieval algorithms have low precision. A desktop interface showing a scrollable grid of many images allows users to rapidly browse the results and pick out relevant images. This paper describes an experiment that compares RSVP (Rapid Serial Visual Presentation) on an iPaq-sized display to the desktop interface. We found that performance was in the same ballpark, but that most users hated the interface.
Mark Derthick. Interactive Visualization of Video Metadata.(Video Demo). in Proceedings of the Joint ACM/IEEE Conference on Digital Libraries, Roanoke, VA, 2001, page 453. Four minute 22MB avi file or 2MB RealMedia.
Much current research on digital libraries focuses on named entity extraction and transformation into structured information. Common types of entities include events, people, and places, which have attributes like birth date or latitude. This video demonstration includes summary views of all 50,000 news segments from CMU’s Informedia Digital Video Library. Drilling down can help find relationships among the entities.
Mark Derthick and Steven F. Roth. Enhancing Data Exploration with a Branching History of User Operations. Knowledge Based Systems, 14(1-2):65-74, March 2001. If you subscribe to this journal, the published version is available from Elsevier ScienceDirect . This is an expanded version of an IUI'00 paper. Accompanying 4 minute video: 37MB avi file or 7MB RealMedia .
Backtracking and investigating alternative scenarios are integral parts of exploratory data analysis. Yet today’s interfaces support only linear undo/redo histories, discarding old scenarios after backtracking and starting a new foray. Further, only one historical state can be visualized at a time, making comparison difficult. In our interface, a scenario tree supports navigating across time and scenarios. Selective undo and redo are accomplished by dragging operations within the tree.
Mark Derthick and Steven F. Roth. Example-based generation of custom data analysis appliances. in Proceedings of Intelligent User Interfaces (IUI '01), Santa Fe, NM, January, 2001, pages 57-64.
Custom scripts or macros speed up repetitive tasks. However, previous attempts to enable non-programmers create custom macros from example task executions have suffered from poor generalization to new situations. This paper suggests that for the limited domain of data analysis, only two simple rules are required for generalization. An observational study suggests that this system represents a useful point on the ease-of-creation vs. expressive power tradeoff appropriate for data analysis.
Joel Welling and Mark Derthick. Visualization of Large Multi-Dimensional Datasets. in Proceedings of Virtual Observatories of the Future 2000. Pasadena, CA. June, 2000. Ed. R. J. Brunner, S. G. Djorgovski, and A. Szalay.
This paper outlines an approach to applying information visualization techniques for interaction with record-oriented multidimensional data to scientific data that includes continuous fields over real 3D space.
Mark Derthick, James Harrison, Andrew Moore, and Steven F. Roth. Efficient Multi-Object Dynamic Query Histograms in Proceedings of the IEEE Information Visualization Symposium (InfoVis), IEEE Press, October 1999, pages 84-91. Also PDF (278K).
In the UIST paper below, visual database queries are made concrete by showing each query graph node as a visualization of the data it contributes. Continuously updating the visualization as the query is changed is a challenge for datasets with hundreds of thousands of records, yet crucial for a direct manipulation interface. This paper was the first to address the efficiency problem for structured queries involving multiple nodes, such as sales and buyers.
Mark Derthick, John
A. Kolojejchick, and Steven F.
Interactive Visual Query Environment for Exploring Data. in Proceedings
of the ACM Symposium on User Interface Software and Technology (UIST '97),
ACM Press, October 1997, pages 189-198. Also PDF
Accompanying 6 minute video: 2MB Shockwave Animation, 7MB avi file, or 3MB RealMedia.
Direct manipulation of visualizations is a powerful technique for performing exploratory data operations such as navigation, aggregation, and filtering. However it does not provide for reuse or modification of exploration sessions. This paper describes a visual language that simultaneously performs data operations and constructs an editable graphical history. These features are supported by the architectural concept of threads, which represent a sequence of navigation steps on particular objects. Because they are tied to particular data objects, they can be directly manipulated. Because they represent operations, they can be generalized into queries.
Mark Derthick, John A. Kolojejchick, and Steven F. Roth. An Interactive Visualization Environment for Data Exploration. in Proceedings of Knowledge Discovery in Databases, AAAI Press, August 1997, pages 2-9. Also PDF (223K)
This paper describes Visage's Visual Query Language and visualization tools, and illustrates their application to several stages of the knowledge discovery process: creating the target dataset, data cleaning and preprocessing, data reduction and projection, and visualization of the reduced data. Unlike previous integrated KDD systems' interfaces, direct manipulation is used pervasively, and the visualizations are more diverse and can be customized automatically as needed. Coordination among all interface objects simplifies iterative modification of decisions at any stage.
I changed fields from AI Knowledge Representation to Information Visualization in 1995. I carried with me an interest in the tradeoff between expressively powerful languages and efficient inference, and also a belief that finding the right data representation is 90% of problem solving.
Mark Derthick. Finding a maximally plausible model of an inconsistent theory. In John A Barnden and Jordan B Pollack, editors, Advances in Connectionist and Neural Computation Theory, volume 1: High-Level Connectionist Models, pages 241-258. Ablex, Norwood, NJ, 1991.
Mark Derthick. A minimal encoding approach to feature discovery. In Proceedings of the Ninth National Conference on Artificial Intelligence, pages 565-571, 1991. Also PDF (2MB).
Mark Derthick. Mundane Reasoning by Parallel Constraint Satisfaction. Research Notes in Artificial Intelligence. Pitman, London, 1990. Reprint of PhD thesis, also available as CMU Technical Report CMU-CS-88-182.
Mark Derthick. Book review of Connections and Symbols. Artificial Intelligence, 43(2):251-265, May 1990.
Mark Derthick. The minimum description length principle applied to feature learning and analogical mapping. Technical Report ACT-AI-234-90, MCC, June 1990. Also PDF (8MB).
Mark Derthick. An epistemological level interface for CYC. Technical Report ACT-CYC-084-90, MCC, February 1990. Also PDF (750KB).
Mark Derthick. Mundane reasoning by settling on a plausible model. Artificial Intelligence, 46(1-2):107-157, November 1990.
Mark Derthick and Joe Tebelskis. `Ensemble' Boltzmann units have collective computational properties like those of Hopfield and Tank neurons. In Proceedings of the IEEE Conference on Neural Information Processing Systems, pages 223-232. American Institute of Physics, 1988.
Mark Derthick. Introduction to mundane reasoning. In David Touretzky, Geoffrey Hinton, and Terrence Sejnowski, editors, Proceedings of the 1988 Connectionist Models Summer School, pages 291-300. Morgan Kaufmann, 1988.
David S. Touretzky and Mark Derthick. Symbol processing in connectionist networks: five properties and two architectures. In Proceedings of IEEE Spring COMPCON87. IEEE, February 1987.
Mark Derthick. A connectionist architecture for representing and reasoning about structured knowledge. In Proceedings of the Ninth Annual Conference of the Cognitive Science Society, pages 131- 142. Lawrence Erlbaum, 1987.
Mark Derthick. Counterfactual reasoning with direct models. In Proceedings of the Sixth National Conference on Artificial Intelligence, pages 346-351, Seattle, Washington, July 1987.
Mark Derthick and David C. Plaut. Is distributed connectionism compatible with the physical symbol system hypothesis? In Proceedings of the 1986 Cognitive Science Conference, pages 639-644. Lawrence Erlbaum, 1986.
Mark Derthick. Variations on the Boltzmann Machine learning algorithm. Technical Report CMU-CS-84-120, Carnegie-Mellon University, Pittsburgh PA, Aug 1984.