CSCE 478/878 Project Ideas

• Text analysis. In particular, extracting data from text and tables in online journal articles. Data would be exported into a database, likely a relational db in Access or SQL. The topic is soil information modified by soil type, agronomic practice, and climate. The system would work interactively with a human user to provide necessary search words and phrases. Has the potential to turn into a supported research position. Talk to me before submitting a proposal on this topic.


• Active learning in text analysis. In particular, experimenting with algorithms from the literature to glean new knowledge out of Civil War-era newspapers and/or from medical research papers. Could result in a co-authorship of a publication. Talk to me before submitting a proposal on this topic.


• Resource scheduling for high-performance computing. Monitor running jobs and review logs to determine if a user is asking for an appropriate amount of resources from the batch scheduler. Talk to me before submitting a proposal on this topic.


• Machine learning in formal concept analysis. Talk to me before submitting a proposal on this topic.


• Budgeted learning, where the labels are free but you have to pay for the attributes; this could be a research project in studying budgeted learning with Bayesian networks


• Active learning, where you have to pay for the labels (report or implementation; could become a research project)


• Optimization techniques in support vector machines (report or implementation; could lead to a research project)


• A survey of kernels for support vector machines (report or implementation)


• Piecewise regression (report and running experiments; could become a research project)


• A study of multiple-instance learning, where each labeled example consists of several instances from the instance space, and we don't know which instance(s) is/are responsible for the label (report, maybe implementation; could become a research project)


• Parallelizing a learning algorithm (implementation; could become a research project)


• Applying learning to biological sequence analysis (implementation; could become a research project)


• Principal Component Analysis (PCA) for feature selection (report or implementation)


• Learning in fuzzy systems, e.g., learning a neural network-based fuzzy controller for a network, a robot, or some other engineering system (report or implementation)


• Learning in network (or other system) control (report or implementation); this may relate a bit to Ying Lu's course on "Self-Managing Computer Systems"


• Automated computer system intrusion detection, i.e. automatically identifing when a person has cracked into a computer system (report or implementation)


• Branch prediction to enhance program performance (computer architecture)


• Learning to classify web pages (report or implementation)


• Learning to make content-based queries in a visual database (report)


• Learning in multi-agent systems (report)


• Using reinforcement learning for optimization problems such as the Traveling Salesman Problem (report, maybe implementation)


• A study of new results in boosting and bagging (report, maybe implementation)


• Using exponentiated gradient in reinforcement learning (report or implementation)


• Using expert-based algorithms (e.g., weighted majority) for pruning decision trees (report or implementation)


• Using expert-based algorithms (e.g., weighted majority) for predicting disk accesses (report or implementation)


• Using expert-based algorithms (e.g., weighted majority) for predicting the stock market or for rating movies (implementation)


• Using statistical hypothesis testing methods for comparing classifiers (report or implementation)


• Dealing with unlabeled data, e.g., clustering, the expectation maximization algorithm, active learning, or co-training (report or implementation)


• Learning hidden Markov models [HMMs] (report or implementation)

Return to the CSCE 478/878 Home Page

Last modified 06 October 2014; please report problems to sscott AT cse.