Refining active learning to increase behavioral coverage

K. Aslam; Y. Luo; R.R.H. Schiffelers; M.G.J. van den Brand

Refining active learning to increase behavioral coverage

K. Aslam, Y. Luo, R.R.H. Schiffelers, M.G.J. van den Brand

Research output: Contribution to conference › Abstract

Abstract

Modern high-tech industry is dealing with the maintenance of complex software systems today which consist of a large number of interconnected software components. Many of these components become legacy over years due to lack of documentation and unavailability of original developers. Several techniques are available in literature to retrieve the behavioral models from the existing software. Among those, the dynamic analysis techniques analyze the actual execution of the software, either via execution traces (passive learning), or by interaction with the software components (active learning). These techniques cannot guarantee alone to learn the complete and correct software behavior due to the limitations of each technique. We present an approach to aid active learning technique with software logs (execution traces) and passive learning result to increase the behavioral coverage of learned models.

Original language	English
Number of pages	1
Publication status	Published - 3 Oct 2018
Event	ACM Celebration of Women in Computing womENcourage 2018 - Belgrade, Serbia Duration: 3 Oct 2018 → 5 Oct 2018 https://womencourage.acm.org/2018/

Conference

Conference	ACM Celebration of Women in Computing womENcourage 2018
Country	Serbia
City	Belgrade
Period	3/10/18 → 5/10/18
Internet address	https://womencourage.acm.org/2018/

Fingerprint

Cite this

Aslam, K., Luo, Y., Schiffelers, R. R. H., & van den Brand, M. G. J. (2018). Refining active learning to increase behavioral coverage. Abstract from ACM Celebration of Women in Computing
womENcourage 2018, Belgrade, Serbia.

@conference{8085f2e6262d484494865965af73a658,

title = "Refining active learning to increase behavioral coverage",

abstract = "Modern high-tech industry is dealing with the maintenance of complex software systems today which consist of a large number of interconnected software components. Many of these components become legacy over years due to lack of documentation and unavailability of original developers. Several techniques are available in literature to retrieve the behavioral models from the existing software. Among those, the dynamic analysis techniques analyze the actual execution of the software, either via execution traces (passive learning), or by interaction with the software components (active learning). These techniques cannot guarantee alone to learn the complete and correct software behavior due to the limitations of each technique. We present an approach to aid active learning technique with software logs (execution traces) and passive learning result to increase the behavioral coverage of learned models.",

author = "K. Aslam and Y. Luo and R.R.H. Schiffelers and {van den Brand}, M.G.J.",

year = "2018",

month = "10",

day = "3",

language = "English",

note = "ACM Celebration of Women in Computing<br/>womENcourage 2018 ; Conference date: 03-10-2018 Through 05-10-2018",

url = "https://womencourage.acm.org/2018/",

}

TY - CONF

T1 - Refining active learning to increase behavioral coverage

AU - Aslam, K.

AU - Luo, Y.

AU - Schiffelers, R.R.H.

AU - van den Brand, M.G.J.

PY - 2018/10/3

Y1 - 2018/10/3

N2 - Modern high-tech industry is dealing with the maintenance of complex software systems today which consist of a large number of interconnected software components. Many of these components become legacy over years due to lack of documentation and unavailability of original developers. Several techniques are available in literature to retrieve the behavioral models from the existing software. Among those, the dynamic analysis techniques analyze the actual execution of the software, either via execution traces (passive learning), or by interaction with the software components (active learning). These techniques cannot guarantee alone to learn the complete and correct software behavior due to the limitations of each technique. We present an approach to aid active learning technique with software logs (execution traces) and passive learning result to increase the behavioral coverage of learned models.

AB - Modern high-tech industry is dealing with the maintenance of complex software systems today which consist of a large number of interconnected software components. Many of these components become legacy over years due to lack of documentation and unavailability of original developers. Several techniques are available in literature to retrieve the behavioral models from the existing software. Among those, the dynamic analysis techniques analyze the actual execution of the software, either via execution traces (passive learning), or by interaction with the software components (active learning). These techniques cannot guarantee alone to learn the complete and correct software behavior due to the limitations of each technique. We present an approach to aid active learning technique with software logs (execution traces) and passive learning result to increase the behavioral coverage of learned models.

M3 - Abstract

ER -