DSpace@RPI
DSpace@RPI is a repository of Rensselaer Polytechnic Institute's theses and dissertations which are available in digital format, largely from 2006 to present, along with other selected resources.
Recent Submissions
Item type:Item, Knowledge Assisted Search(JSAI, 1997-08-23)Web application demands for more intelligent search functionality provide many opportunities for applications of artificial intelligence techniques. Websites are encompassing more complex content and functionality and are becoming more difficult to search by standard web search engines (e.g., Alta Vista[12], Lycos[14], etc.). Even today, many sites need to provide their own search functionality in order to (i) limit the scope of the search, (ii) exploit knowledge of the structure of the internal site, and (iii) exploit understanding of the content of the site. Many of these searches are beginning to use artificial intelligence techniques to provide better organization of, understanding of, access to, and reasoning with information. In this paper, we will present our vision of a "smart search" functionality which relies on a standard text search augmented with knowledge representation techniques. We I will also describe directions for further work.Item type:Item, Automatic Re-Programming for Robustness(IEEE, 1986-12-01)Even rigorously verified programs may be poorly protected against actual hazards encountered in large embedded systems (e.g., corrupted data, unforssen user actions, or peripheral device failures). RIP is a knowledge-based automatic programming tool to retrofit correct but "naively" written algorithms with enhanced error handling properties, including error detection and damage confinement.Item type:Item, Integrating Astrobiology Research into Undergraduate Curricula: From early Earth to Mars(2026-05-05)Two innovative educational programs at RPI's Rensselaer Astrobiology Research & Education (RARE) Center are bringing authentic astrobiology research into undergraduate classrooms, designed for students with different interests and skills. The LEGOS targets experimental laboratory work. Students choose a specific organic molecule that may have been important for early life, then design experiments to test its stability under different planetary conditions. They collect data that contributes to a growing database helping scientists understand how life's building blocks survived on early Earth. Students experience real scientific challenges including data analysis, uncertainty, and interpretation. The Mars Visualization Project engages computer science and data analysis students. Working in teams, they analyze data from NASA's Mars 2020 rover, learning to work with data from different analytical instruments. Students develop interactive applications that help scientists explore and understand Mars data, creating tools that are the first to combine observations from multiple rover instruments. Both programs emphasize collaboration, dealing with uncertainty, and making scientific judgments rather than more traditional, prescribed, and often well-tested research projects. They demonstrate how astrobiology can welcome students from diverse backgrounds into research, helping build an inclusive and interconnected scientific community while preparing students for real-world scientific work.Item type:Item, Unsupervised: Pitfalls and Best Practices in Dataset Curation for Machine Learning in Astrobiology(2026-05-05)Machine learning and artificial intelligence are becoming important tools in astrobiology. However, these tools are only as good as the data they use. This contribution examines how to properly organize and prepare scientific data for machine learning applications in astrobiology.Four specific use cases that highlight best practices and also illustrate potential pitfalls: analyzing scientific publications to understand research and community trends; data curation and analysis of decades of prebiotic chemistry experiments; assembling, analyzing and visualizing various datasets from NASA's Mars rover; and comparing complex chemical mixtures that result from organic transformation experiments.The key findings show that careful, thoughtful data organization is essential. Best practices include testing datasets multiple times, using standardized classification systems, combining different types of measurements, and rigorously checking how data is processed. Without these practices, machine learning can produce misleading results that confirm existing biases rather than revealing new insights.Rather than treating AI as a tool to re-analyze historical data, we argue that datasets should be designed from the start with AI compatibility in mind. This approach can accelerate scientific discovery while making research more transparent, reproducible, and useful across the astrobiology community.Item type:Item, Formalizing the semantics of sea ice(Springer Nature, 2014-03-01)We have initiated a project aimed at enhancing interdisciplinary understanding and usability of polar data by diverse communities. We have produced computer- and human-understandable models of sea ice that can be used to support the interoperability of a wide range of sea ice data. This has the potential to improve scientific predictive analyses and increase usage of the data by scientists, modelers, and forecasters as well as residents of communities that rely on sea ice. We have developed a family of ontologies, leveraging existing best in class models, including one module describing physical characteristics of sea ice, another describing sea ice charts, and a third modeling “egg codes” - an internationally accepted standard for symbolically representing sea ice within geographic regions. We used a semantic Web methodology to rapidly gather and refine requirements, design and iterate over the ontologies, and to evaluate the ontologies with respect to the use cases. We gathered requirements from a wide range of potential stakeholders reflecting the interests of operational ice centers, ice researchers, and indigenous people. We introduce the driving use case and provide an overview of the resulting open source ontologies. We also introduce some key technical considerations including the prominent role of provenance, terms of use, and credit in the model. We describe how the ontologies are being employed and highlight their compatibility with a wide range of existing standards previously developed by many of the stakeholder communities.
Communities in DSpace@RPI
Select a community to browse its collections.