223 lines
8.4 KiB
Markdown
223 lines
8.4 KiB
Markdown
file:: [MODELS23-DS_paper_11_1690904186006_0.pdf](../assets/MODELS23-DS_paper_11_1690904186006_0.pdf)
|
||
file-path:: ../assets/MODELS23-DS_paper_11_1690904186006_0.pdf
|
||
|
||
- software engineering involves implementing an executable program which principally includes writing code.
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c928f1-b5d1-4ef3-8a53-8479e4c55647
|
||
- managing data and training models
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c928fd-096c-4430-9d5c-769f186aac3a
|
||
- elatively less emphasis on the code writing
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92910-7bdb-44c2-8cdd-6bc7b977d842
|
||
hl-stamp:: 1690904850983
|
||
- how to improve the engineering of systems with AI components (SE4AI
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92922-edb7-45ed-b6f7-c001b8190a3a
|
||
- continuous learning and adaptation
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92933-20d1-4b68-a06e-89eb43f559b8
|
||
- ade-offs between software process modeling languages in AI/ML context must be re-examined
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: yellow
|
||
id:: 64c92944-7620-46e8-82a7-10994cdfe2b9
|
||
- In this paper, we propose a framework for modeling and executing software engineering process for AI-enabled system
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92956-2066-4199-a3d7-8e7e32b53c03
|
||
- Software process model
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c92b40-e90e-42ec-8a6d-b9f09927123d
|
||
- structured view of the processes, facilitating improvements, and permitting of the processes standardization and reu
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c92b4c-7ad7-45c0-bfc3-73852cd3d537
|
||
- activity-centric
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c92b5d-4413-4791-8250-c14b1c11189b
|
||
- artifact-centric
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: green
|
||
id:: 64c92b63-a12a-4b1d-8aae-a9461b6c2b29
|
||
- Artificial Intelligence-enabled Systems (AIS) are used increasingly in real-world applications. This is mainly due to the success of deep learning algorithms in the fields of image processing, speech recognition and machine translation
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92bc5-03bd-47a2-a429-fc5eee0ef973
|
||
- AIS is a software system including one or more ML components.
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92bd4-c5ff-4ef6-bf5b-d716404a359e
|
||
- his changes
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: red
|
||
id:: 64c92be9-44a2-479f-bb8b-78293d0d1f70
|
||
- new demands changes in the development process aros
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: red
|
||
id:: 64c92bf9-70db-441e-ae0c-7b69b70ec7f1
|
||
- Q1: What challenges emerge in the management of software engineering processes for AI-enabled systems
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64c92c23-013d-4d2a-928f-0cfec5691230
|
||
- An Artificial Intelligence-Enabled System (AIS) is a software system in which at least one of its components relies on one or more ML models
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: purple
|
||
id:: 64c92c7f-9de7-4bfc-88aa-d04a974419d5
|
||
hl-stamp:: 1692254584445
|
||
- ML model can be defined as ”a trained instance of a specific machine learning algorithm” [6].
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: purple
|
||
id:: 64c92c8d-f876-4b69-b263-9dcff892b6f1
|
||
hl-stamp:: 1692254587588
|
||
- Managing Data and Training ML model [
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c92f21-b214-4394-82a4-3009cdb64852
|
||
- In the development of AIS, it is essential not only to define the requirements for the final product but also to establish the requirements for the learning process
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: blue
|
||
id:: 64c9311c-47e0-46ab-b443-615081e46d2d
|
||
hl-stamp:: 1692254744355
|
||
- ence, it can be stated that the effective satisfaction of NFRs becomes critic
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c932d2-6a74-4aaa-973f-6a0647afd612
|
||
- Software Testing (ST) is the process of verifying that a software behaves as specified, detecting errors, and validating that what has been specified is what the stakeholders actually needed [12]
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c932f6-9509-4030-8e23-1b658fbd1e6c
|
||
- Input-level Testing
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c93311-5d7a-483a-8896-708aec8a40fe
|
||
- Model-level Testin
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c93316-7cd6-49e2-853e-072a855914bc
|
||
- ntegration Testing
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c9331e-2631-4ed8-9b44-d4ff2e32515c
|
||
- System Testing
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c93322-7155-44ce-a5ce-f540288cc1c7
|
||
- heir primary purpose distinguish them from each other.
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c93450-86d4-4840-b1a8-9490ccd8e3c4
|
||
- maintaining of the Traceability. Traceability imposes the conditions that the interdependencies among the artifacts be made explicit and that each artifact be trackable longitudinally through the entire development process.
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64c93460-8516-4f71-8aba-427db12bc628
|
||
- We provide partial model of ML workflow practiced by Microsoft team. Microsoft’ ML workflow is nine stages with some data-oriented stages (e.g., data cleaning, and data labeling) and others model-oriented(e.g., feature engineering, and model training) [2]
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: purple
|
||
id:: 64c93490-b68c-49a0-b42e-82e1ec896e60
|
||
- (2) Artifacts of novel types are emerged
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: green
|
||
id:: 64c93526-d05f-42e1-8c1b-b9f6ddedeb7a
|
||
- (1) Joint effort of roles from different backgrounds involved in the development process
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: green
|
||
id:: 64c9352e-2c1f-4d82-9f66-1ed59a6754a4
|
||
- Compatibility ensures that the ML components necessary for achieving reproducibility work together without conflicts
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: green
|
||
id:: 64c93574-9723-481c-b6b0-d71d1d042589
|
||
- complex
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: yellow
|
||
id:: 64c93642-2f81-4249-a1b4-3ddd616cf7d9
|
||
- effectively model SE processes for AIS
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: green
|
||
id:: 64caccc4-bd9d-44a6-9dca-e7cec9e57a39
|
||
- the same set of dependencies and versions are used consistently across different platforms and collaborators)
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: purple
|
||
id:: 64cacd41-b191-49d1-89ad-fabee626a9a0
|
||
- refinement of the ML model to achieve the desired performanc
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: purple
|
||
id:: 64cacdb0-8843-4df6-9efb-90b89656d54b
|
||
- the need to revisit the data management, feature engineering, and model training steps multiple times to experiment with different approaches and refine the ML solution.
|
||
ls-type:: annotation
|
||
hl-page:: 3
|
||
hl-color:: green
|
||
id:: 64cacdc1-5647-42b8-92fc-9bf0fd453298
|
||
- r,
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: red
|
||
id:: 64ddc115-d113-4060-9906-c2e9808d39e6
|
||
- Q2: What are the essential characteristics that process modeling should encompass to address the novel challenges introduced by AI-based systems development processes?
|
||
ls-type:: annotation
|
||
hl-page:: 1
|
||
hl-color:: purple
|
||
id:: 64ddc160-cb14-4de4-a7c7-884bfef31415
|
||
- demands change
|
||
ls-type:: annotation
|
||
hl-page:: 2
|
||
hl-color:: red
|
||
id:: 64ddc1ea-a881-471e-ba0c-7004ee7bdbc8
|
||
- (1) Define Domain-specific modeling (DSL) for creating models that are specific to SE for AIS (including the specific related AI tasks, roles, and artifacts). Defining these concepts proves beneficial for enhancing communication among different participants, automating trace link generation, enabling more flexible execution, and facilitating partial process modeling.
|
||
ls-type:: annotation
|
||
hl-page:: 4
|
||
hl-color:: yellow
|
||
id:: 64ddc444-165f-485d-bd43-f80a5a692e46
|
||
hl-stamp:: 1692255302067
|
||
- We address this lack by proposing a goal-oriented ACPM methodology for systematically modeling the processes [15].
|
||
ls-type:: annotation
|
||
hl-page:: 4
|
||
hl-color:: yellow
|
||
id:: 64ddc4ef-3de7-4836-8994-3d0066c252f3
|
||
- (3) Develop a data-driven process engine capable of executing fragmented artifact process models, and generating a macro model derived from the partially modeled process.
|
||
ls-type:: annotation
|
||
hl-page:: 4
|
||
hl-color:: yellow
|
||
id:: 64ddc53c-0a3a-4590-987d-5b594e458d7d |