Software Engineering Methodology

Running Head: SOFTWARE METHODS
0
Software Engineering
Artificial Neural Networks in Software Engineering
(Student Details: )
12/18/2019

SE
1
Contents
Research Report: Software Engineering Methodology..............................................................2
Introduction................................................................................................................................2
Deep Learning (DL) in Automotive Software...........................................................................2
Conclusion..................................................................................................................................5
References..................................................................................................................................6

SE
2
Research Report: Software Engineering Methodology
Introduction
This research report is based on software engineering methodology with the help of artificial
neural networks. Here, the paper is going to explore how artificial neural networks’
application is done in the field of software engineering. While writing this report, a journal
article will be critiqued which is based on the application of neural networks into software
engineering 1. In this context, software engineering is completely new as well as the ever-
changing field. In this way, the challenge of meeting strict project schedules with modern
software always needs that the software engineering field is automated largely. In addition,
with the help of Artificial Neural Networks (ANN), human efforts and intervention can easily
be minimized to an optimum level 2. In this era, researchers are exploring the potential of
machine learning methods. It is because they are highly adaptable and have learning abilities
as well as non-parametric. Throughout this report, the role of ANN can be effectively utilised
for building different tools for software development and thereby maintenance functions.
Deep Learning (DL) in Automotive Software
Based on the selected article, it can be said that artificial neural networks can be used in
software engineering 2. In general, ANNs are those computing systems that are typically
stimulated from the biological neural networks which use animals’ brains. It is the fact that
ANNs learn with the help of examples so that they can perform tasks and thereby they are
generally without being programmed with the function-specific rules. The article is all about
applying deep learning into software engineering and automotive software 2. Hence,
understanding deep learning is important. As we know, deep learning is very exciting as well
as a powerful branch of the subject machine learning. In addition, this is a technique that
teaches machines to do what comes naturally to humans with the help of daily life examples
3. In this context, DL is a key part of a wider family of ML (referred to ML) machine
learning methodologies on the basis of ANNs. In other words, DL is also known as
hierarchical as well as deep structured learning. In this way, based on the chosen article it can
be said that ANNs are typically inspired by distributed communication nodes and information
processing within biological systems 4.

SE
3
Moreover, DL is highly useful for software engineering as networks driven by DL are highly
useful. DL networks contain the number of node-layers with which data can be passed into a
multi-step procedure of pattern recognition and many more 4. The authors have suggested
that whenever more than three layers are there containing input as well as output, hence it is
known as deep learning 2. In this context, some key DL models are used for a variety of
complicated tasks as follows:
 Autoencoders for recommendation systems
 ANN for classification and regression
 Self-organizing maps for the purpose of feature extraction
 Convolution Neural Networks (CNN) for computer vision
 Time series analysis through recurrent neural networks
In this chosen article, authors have covered everything related to ANN in software
engineering 2. Specifically, this chosen study has focused on ANN application in the form of
DL for automotive software. The modern researchers have launched a framework which
actually supports disciplined and robust development lifecycle for automotive software 5.
Based on this study, it can be said that DL is emerging as crucial to the development of
automotive software for apps like autonomous driving.
In addition, there can be great automotive implementations of DL in various forms 2. For
example, Google has made remarkable as well as highly visible investments within
autonomous-vehicle development. The automotive software-driven by DL is helpful in
detecting pedestrians into difficult and challenging scenarios. The article has found that DL
systems have attained great performance while making the error rate for machine vision less
than humans 4. Apart from this, the concept of basic deep neural networks (DNN) has been
introduced with the help of this research study. It has been observed that DL excels at finding
patterns whenever the input is an enormous analog data rather than the small number of audio
data and image data. While analysis this paper, it has been found that DNNs structure is
greatly flexible and hence can be customized with the help of a selection of key factors like
the number of units per layer, hidden layers, and connections per unit. The authors suggested
that such attributes are known as hyperparameters and also determine the DL based system’s
behaviour and structure 2.

SE
4
On the other hand, the evaluation part of the chosen study is suggesting that DL has 4 major
characteristics as follows:
 Interconnected elements like nodes or neurons which are nonlinear computational
elements 3.
 I/O mapping which is typically created with the learning procedure
 Adaptability to environmental or ecological changes
 Fault tolerance 6.
In this way, the article suggests that DNNs’ processing ability is stored within connection
weighs, which are acquired with adaption towards a training pattern set 5. If we talk about the
critique part, then it has been realised that key examples are missing from the study. The
examples which could have explained the theory to practice in relation to automotive
software is missing 1. In the context of training a DNN, training typically functions as a
programming activity. It has been explained with an example of computer vision. Generally,
DNN training needs a large set of images for matching the input to the statistically expected
correct outcomes which are featuring colours, edges, and shapes 7. Moreover, as per the
discussion that took place in the chosen article, DL based systems are emerging pervasive in
automotive software. In addition, within automotive software engineering communities,
awareness is also increasing in integrating DL based development with conventional
development methods 7.
In addition to that, DNN training by using ad hoc training data has been shown in this article.
The research is showing that DNN training is the key as this allows whole exploitation of the
DNN’s capability for detecting an object while taking into account the selected image’s
context. Besides, the role of ANN is clear with the application of DL in automotive software
i.e. DNN and it is pivotal in the software development for vehicle functions that rely upon DL
8. Furthermore, automotive applications considerably integrate supervised training with the
help of reinforcement training 9. In general, this article has talked about the choice of training
strategy which depends on the DNN type and the key issues under consideration. Throughout
the article, a lifecycle for DL has been explored with the W model. Here, the W model
conceptually combines a V model for data development with the V model for software
development 10. However, DL intrinsically introduces attributes to software development
that do not wholly fit the V model. Based on the analysis of the chosen article, it can be said
that the valuable aspect of AI in automotive apps is that the use of ANN in software

SE
5
development is continuously learning as well as adjusting the major rules as it often uses for
navigating the roads. In this context, data’s key role makes the introduction of the W model
essential for the software developers. As we know, some key trends are majorly affecting
automotive software development which is as follows:
 Functional safety 11.
 A hike in connectivity needs
 Consolidation of ECU functionality
 An increase in security needs
 OTA updates
 Leveraging OSS (open-source software)
 Long-tern support needs 12.
Conclusion
In whole, this research report has successfully reviewed, explained and explored the chosen
article covering ‘deep learning in automotive software’. In addition, this paper has covered
different aspects of ANN when used in software engineering. While analysing the article, it
has been found that DL uses data as well as algorithms for imitating the cognitive activities of
the human mind. Apart from this, the evaluation of the article has covered both the positives
and negatives of the chosen article. The ability of learning and solving issues has been
explored with the help of an application of ANN in automotive software. It has been found
that the entire automotive industry is at the front of utilising ANN to augment, mimic and
thereby assist the human acts while also leveraging the advanced reaction times and point out
the precision of machine-based systems. Therefore, ANNs are getting used for autonomous
and semi-autonomous vehicles of the upcoming era as they will depend on AI systems.

SE
6
References
x
[1] P. Stone, R. Brooks, R. Calo, and G. Hager, "Artificial intelligence and life in 2030,"
One Hundred Year Study on Artificial Intelligence: Report of the 2015-2016 Study
Panel, p. 52, 2016.
[2] F. Falcini, G. Lami, and A. Costanza, "Deep learning in automotive software," IEEE
Software, vol. 34, no. 3, pp. 56-63, 2017.
[3] B. Sinha, D. Sinhal, and B. Verma, "A software measurement using artificial neural
network and support vector machine," International Journal of Software Engineering
and Applications, vol. 4, no. 4, 2013.
[4] Tractica. (2019) Tractica. [Online]. https://www.tractica.com/research/artificial-
intelligence-for-automotive-applications/
[5] Ann Lewis. (2018) Medium. [Online]. https://medium.com/@ann_lewis/the-software-
engineering-learning-plan-c4d97aedf913
[6] C. S. Krishnamoorthy and S. Rajeev, Artificial intelligence and expert systems for
engineers. US: CRC Press, 2018.
[7] M. White, C. Vendome, M. Linares-Vasquez, and D. Poshyvanyk, "Toward deep
learning software repositories," in 12th Working Conference on Mining Software
Repositories, 2015, pp. 334-345.
[8] Y. Duan and J. S. Edwards, "Artificial intelligence for decision making in the era of Big
Data–evolution, challenges and research agenda.," International Journal of Information
Management, vol. 48, no. 1, pp. 63-71, 2019.
[9] LYUDMYLA NOVOSILSKA. (2018) Ignite. [Online].
https://igniteoutsourcing.com/automotive/artificial-intelligence-in-automotive-industry/
[10] C. Dirican, "The impacts of robotics, artificial intelligence on business and economics,"
Procedia-Social and Behavioural Sciences, vol. 195, pp. 564-573, 2015.
[11] J. De Gea, J. Nicolas, J. Aleman, and A. Toval, "Requirements engineering tools:
Capabilities, survey and assessment," Information and Software Technology, vol. 54, no.
10, pp. 1142-1157, 2012.
[12] M. Bisi and N. Goyal, Artificial Neural Network Applications for Software Reliability
Prediction. UK: John Wiley & Sons, 2017.
x

SE
7