Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


thesis_man_bhujel_latest.pdf (1.15 MB)

ETD Abstract Container

Abstract Header

Performance Enhancement of Data Retrieval from Episodic Memory in Soar Architecture

BHUJEL, MAN BAHADUR
http://rave.ohiolink.edu/etdc/view?acc_num=toledo1524837171020967

Abstract Details

2018, Master of Science, University of Toledo, Electrical Engineering.
Episodic memory has been the key component of various intelligent and cognitive architecture that includes the autobiographical events of past experiences. The implementation of episodic memory enhances the performance of cognitive agents by utilizing past history for decision making. During episodic retrieval in Soar architecture, the cue matching step involves a two stage process to improve the performance of the architecture. Soar implements cue matching as a surface cue analysis, which fi nds candidate episodes based on the matched leaf node. It then performs a structural match on candidate episodes with full surface match. However, continuous design research is still needed for minimizing the operational time of episodic processes along with timely cue-matching as episodic memory grows. This thesis provides an insight to improve on both stages of cue matching, which ultimately leads to a quick retrieval of episodes. First, the approximations of original implementation base-level activation (BLA) is implemented for determining the feature weight for surface cue analysis. These methods are computationally efficient. Second, a new approach of solving the constraint satisfaction problem (CSP), arc-consistency algorithm is implemented for the structural cue analysis. For the experiment, two of the most frequently used testing environment, Eaters and TankSoar, are chosen. From the first experiment, it is found that the Eaters agent provides comparable performance with approximations of BLA and demonstrate applications of approximations. The approximation of BLA has high computational efficiency. Determining activation value of working memory elements (WMEs) is a part of cue matching; hence, incorporating approximation of BLA leads to faster retrieval of episodic memory. Also, using a new technique for structural graph match, this thesis obtains less query time compared to original one. This also leads to decrease in retrieval time. Moreover, the results show that as the size of episodic memory increases, the rate of change of retrieval time with arc-consistency decreases. With these results, the ultimate goal of the research is achieved.
Vijay Devabhaktuni (Committee Chair)
Ahmad Javaid (Committee Co-Chair)
Devinder Kaur (Committee Member)
88 p.
Electrical Engineering
Soar Architecture; Episodic Retrieval; Base-level Activation; Petrovs Approximation; Structural Graph Matching; Backtracking; Arc-Consistency; Cue-matching

Recommended Citations

Citations

  • BHUJEL, M. B. (2018). Performance Enhancement of Data Retrieval from Episodic Memory in Soar Architecture [Master's thesis, University of Toledo]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1524837171020967

    APA Style (7th edition)

  • BHUJEL, MAN. Performance Enhancement of Data Retrieval from Episodic Memory in Soar Architecture. 2018. University of Toledo, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=toledo1524837171020967.

    MLA Style (8th edition)

  • BHUJEL, MAN. "Performance Enhancement of Data Retrieval from Episodic Memory in Soar Architecture." Master's thesis, University of Toledo, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1524837171020967

    Chicago Manual of Style (17th edition)

toledo1524837171020967
432
© 2018, some rights reserved.Creative Commons License Performance Enhancement of Data Retrieval from Episodic Memory in Soar Architecture by MAN BAHADUR BHUJEL is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by University of Toledo and OhioLINK.