User Tools

Site Tools


7a.016.DU - A Continual Learning Framework for Domain Adaptation and Provenance Tracking

Project - Summary

  • Unordered List ItemImplement a continual learning framework that can incrementally update classifiers and efficiently relabel old datasets without the cost of reprocessing the existing database. Significance: In previous solution, the classifiers are entirely retrained when the database gets updated by new information – this is redundant and wastes time and computation. Such a framework can greatly reduce the computational redundancies and make instant decisions using latest information.
  • Unordered List ItemEnable novel classes detection. Significance: One would like to detect novel classes or abnormal data during the classification process. This will enable one to classify on a large volume data, which occurs in healthcare, social networks.
  • Unordered List ItemDevelop Antibiotic Resistance Prediction model that learns from whole genome sequencing data and predict the MIC (minimum inhibitory concentration) rate and find important genes that are predictive or associated with the AR phenotype. Significance: We can predict the MIC from the whole genome sequence which can be helpful in clinical applications. In addition, the important genes identified by this model can be used as sequencing target instead of performing a whole genome sequencing to reduce the cost.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Gail Rosen PI (215) 895-0400 Drexel University
Zhengqiao (Joe) Zhao Student Not available Not available Drexel University
Taha ValizadehAslani Student Not available Not available Drexel University
Creighton Kagey Project Mentor Not available Becton Dickinson (BD)
Funded by: Becton, Dickinson & Company

Project - Novelty of Approach

Most provenance must be manually tracked and classifiers do not have the ability to adapt to new environments/domains.

Project - Deliverables

1 A continual learning prototype software that ingests, and classifies digital informational objects like DNA sequences
2 Publications/patent/documentations that discusses the implementation and validation of the framework
3 Project report on potential findings/discoveries in the database using the proposed algorithm
4 An antibiotic resistant model that predict the MIC for different bacteria
5 A list of important genes identified by the model and some interpretation and project report on potential findings/discoveries in the database using the proposed algorithm

Project - Benefits to IAB

We are preparing a journal article with title “NBC++, an incremental implementation of the Naïve Bayes Classifier for metagenomic taxonomic classification” for this project (this report is a subset of the paper that contains the most relevant information) and will also publish our source code on Github.

Project - Presentation Video

Project - Documents

FilenameFilesizeLast modified
7a.016.du_mid-year_report.docx239.8 KiB2019/12/20 10:50
7a.016.du_ip_info_sheet.docx19.3 KiB2019/08/19 12:47
7a.016.du_final_report.docx3.7 MiB2019/08/19 12:43
7a.016.du_poster_2018_spring_meeting.pptx1.6 MiB2019/08/13 15:02
7a.016.du_2018_fall_meeting_poster.pptx202.4 KiB2019/08/13 15:02
7a.016.du_cvdi_mid-year_report_7a.016.du.docx237.9 KiB2019/08/13 15:02
7a.016.du_executive_summary.docx52.8 KiB2019/08/13 15:02
7a.016.du_confluence_project_page.pdf141.6 KiB2019/08/13 15:02
7a.016.du_quad_chart_2018_spring_meeting.pptx941.7 KiB2019/08/13 15:02
projects/year7/7a.016.du.txt · Last modified: 2020/01/21 12:39 by sally.johnson