User Tools

Site Tools


projects:year8:8a.004.uva

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
projects:year8:8a.004.uva [2019/09/25 16:59]
sally.johnson created
projects:year8:8a.004.uva [2019/10/21 17:12] (current)
matt.delcambre
Line 2: Line 2:
 ===== Project - Summary ===== ===== Project - Summary =====
 <WRAP leftalign box > <WRAP leftalign box >
-This project aims to improve data-driven decision-making models for an autonomous systemparticularly with the goal of making explainable predictions for intent of other surrounding systems without any direct communicationThe goal of this project is to improve the usability of and trust in such data-driven decision-making models by integrating human experts’ decisions and feedback into the model’s knowledge poolThe project will create a model that jointly reasons about and predicts a distribution over plausible decisions and uses human expert knowledge to generate an optimized decisionThe application will focus on autonomous vessels at sea, with the goal of broad applicability to other domains.+  * Much of the recent advancement in machine learning has been through neural networks (NNs), and in particular deep learning (DL) has shown record-breaking performance across many domains 
 +  * In some fields where there is a human element involvede.g. in medicine and finance, there can be legal or ethical considerations that require decisions to be interpretable/explainable. This requirement often prohibits the use of NNs and other so-called 'black-box' models
 +  * There have been some proposed techniques to help explain the internal workings of NNs, many focused on image/visual recognition. 
 +  * One of these methods known as feature occlusion shows promise for non-image and multi-modal data. 
 +  * We propose to investigate the use of feature occlusion and potentially develop other techniques for use as general methods for explaining and interpreting deep learning models.
  
  
 </WRAP> </WRAP>
 ===== Project - Team ===== ===== Project - Team =====
-^ Team Member   ^ Role  ^ Email               ^ Phone Number    ^ Academic Site/IAB       ^ +^ Team Member    ^ Role                ^ Email                ^ Phone Number    ^ Academic Site/IAB       ^ 
-Cody Fleming  | PI    cf5eg@virginia.edu  | (434924-7460  | University of Virginia +Peter Beling   | PI                  pb3a@virginia.edu    | (804) 982-2066  | University of Virginia 
-                    |                                     | Funded by: **Leidos**   |+| Stephen Adams  | Research Scientist  | sca2c@virginia.edu   | (757870-4954  | University of Virginia 
 +Alex Langevin  Student             | arl4zk@virginia.edu  | Not available   | University of Virginia 
 +|                |                     |                      |                 | Funded by: **U-ANGEL**  |
 ===== Project - Novelty of Approach ===== ===== Project - Novelty of Approach =====
 <WRAP leftalign box > <WRAP leftalign box >
  
-The usability of data-driven decision-making solutions in safety-critical applications is limited by the absence of human-like inference and reasoning about decisions. Moreover, most models overlook the possibility of more than one feasible decision and end up optimizing on average behavior. This project is expected to improve decision-making models by incorporating human expert knowledge and interpret ability in the model. The project will focus on maritime domain that has different conditions from land or air based research projects focusing on autonomous agents.+This project seeks to determine the feasibility and performance of feature occlusion and other techniques as general methods of explainable AI with test case implementation in IAB member defined settings
  
  
Line 19: Line 25:
 ===== Project - Deliverables ====== ===== Project - Deliverables ======
  
-^    ^ Deliverables                       +^    ^ Deliverables                                                                    
-| 1  | Briefing of model design           +| 1  | Source code                                                                     
-| 2  | Literature review document         | +| 2  | Report of results and list of significant predictive features and interactions  |
-| 3  | Software and associated artifacts  +
-| 4  | Final report                       |+
 ===== Project - Benefits to IAB ===== ===== Project - Benefits to IAB =====
 <WRAP leftalign box > <WRAP leftalign box >
    
-This project will directly inform technology development for autonomous systems. It will benefit any IAB member that is developing autonomous agents.+  * Improved model accuracy while maintaining interpretability 
 +  * Ability to explain model results where legally or ethically required 
 +  * In commercial or engineering settings, improved business intelligence and diagnostic data gathering
  
 </WRAP> </WRAP>
Line 33: Line 39:
 ===== Project - Documents ===== ===== Project - Documents =====
  
-{{filelist>*8a.005.uva*&sort=mtime&style=table&tableheader=1&showdate=1&showsize=1}}+{{filelist>*8a.004.uva*&sort=mtime&style=table&tableheader=1&showdate=1&showsize=1}} 
 + 
 +===== Life Form Feedback ===== 
 + 
 +<doodle 
 +  title="Year 8 Project Poster Session Feedback (Fall 2019)" 
 +  auth="user" 
 +  adminUsers="matt.delcambre|sally.johnson" 
 +  adminGroups="admin" 
 +  voteType="single" 
 +  fieldwidth="auto|123px" 
 +  closed=“2019-11-05 14:00:00" 
 +  printUser="fullname" 
 +  showMode="all" 
 +  showSum="true" 
 +  userlist="vertical" 
 +
 +   * Great Progress  
 +   * On Course 
 +   * Needs Change 
 +   * Off Course 
 +   * Abstain 
 +</doodle> 
  
 ~~DISCUSSION~~ ~~DISCUSSION~~
projects/year8/8a.004.uva.1569448791.txt.gz · Last modified: 2019/09/25 16:59 by sally.johnson