Name: Zainab Sayyed Roll no.:13327, A.Y.
2023-24
Sem-I
INTRODUCTION
Urban migration, projected to reach over 68% by 2050, strains cities, necessitating smarter solutions
to address congestion, crime rates, and resource management. Smart cities, built on technological
advancements, aim for sustainability, efficiency, and improved living conditions.
Data from sensors, cameras, and social media create a digital ecosystem processed via communication
systems. Challenges include fragmented AI and analytics, human involvement, and the need for adaptable
machine learning techniques.
To tackle these issues, this paper proposes a self-building AI framework based on the Growing Self-
Organizing Map (GSOM) algorithm. GSOM enables self-learning without pre-labeled data, reducing human
intervention in real-time anomaly detection and trend prediction. The Distributed GSOM further provides a
Global Position Map for integrated processing.
The paper emphasizes the importance of bringing technology into practical applications for better urban
living and sustainability in cities.
The topic of the paper is "Self-Building Artificial Intelligence and Machine Learning in the Smart City
Environment." The paper focuses on the application of AI and machine learning techniques in smart cities, with
a specific emphasis on the use of unsupervised machine learning technique called Self Organizing Map (SOM)
and a suite of self-building (structure-adapting) versions of the SOM. The authors propose a framework that
utilizes cloud computing platforms and local/global processing to enable self-building AI capabilities in smart
cities.
The paper provides a background on smart cities, cloud computing platforms for smart cities, and the use of AI
within smart cities. It highlights the limitations of traditional AI approaches in handling the complexities and
dynamics of data in smart city environments. The proposed framework aims to overcome these limitations by
creating AI structures that can adapt and evolve incrementally based on the specific characteristics of the data.
The practical implications of the proposed framework are also discussed, including its potential applications in
traffic management, emergency response, public services optimization, and infrastructure planning. The paper
concludes with a discussion on the applicability and effectiveness of the proposed framework in real-world
situations, emphasizing the benefits of self-building AI in handling diverse and complex data sources.
Overall, the paper presents a comprehensive exploration of the concept of self-building AI in the context of
smart cities, offering insights into its theoretical foundations and practical implications.
1
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
LITERATURE SURVEY
Sr. Year Authors Title Publicat Technology Limitations
ion
no
Damminda -Made a -Job
-Self-
significant displacements.
Alahakoon Structuring
contribution with -Ethical concerns.
1 AI.
2018 international -A lack of human-
impact towards -Cognitive like creativity and
the computing Empathy.
advancement of
Artificial -Deep
Intelligence Learning
-Applied
AI
research.
-city -Technical
Yan Xu -Associate background
sustainabilit
Director of the needed.
y and waste
International
managemen
Research
t.
2 2019 Institute for
-multi
sustainable
criteria
Operations at
decision
Northwestern
analysis.
Polytechnical
university
2
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
complex
technology
involved
-Associate -Information
Daswin Editor of the fusion
De Silva IEEE Autonomous
Transactions learning
on Industrial
Informatics
-Natural
3 Language
2019
-Processing
Incremental
Learning
-Professor of
Uthayasankar -Block
Technology
Sivarajah chain use
Management
in
and Circular
Financial
Economy at
services
the School of
and Smart
Management,
cities.
4 University of
2021 Bradford
3
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Techniques:
A Self-Organizing Map (SOM), also known as a Kohonen map, is a type of artificial neural
network used for data visualization, clustering, and dimensionality reduction. It was developed by Teuvo
Kohonen in the 1980s. Here's a detailed explanation of how SOM works:
1. Network Structure: A SOM consists of a grid of nodes or neurons organized in a two-
dimensional (usually) lattice. Each node represents a prototype or a cluster center.
2. Input Data: The input data can be multi-dimensional, typically numerical, and is used to train the
SOM. Each input vector corresponds to a point in the high-dimensional space.
3. Initialization: Initially, the weights of the neurons are assigned randomly or using some other
initialization method. These weights are the same dimension as the input data.
Training Process:
1. Competition Phase: For each input vector, the SOM identifies the best-matching neuron,
known as the "winner." This is done by calculating the similarity between the input vector and
the neuron's weights. The neuron with the closest weights to the input vector wins.
2. Cooperation Phase: Neurons in the neighborhood of the winning neuron are also updated to
become more similar to the input vector. This encourages nearby neurons to respond to
similar input patterns, leading to spatial organization in the SOM.
3. Adaptation: The winning neuron's weights are adjusted to become more similar to the input
data. This process continues for each input vector, gradually adjusting the neuron weights.
4. Learning Rate and Neighborhood Radius: Two important parameters in SOM training are
the learning rate and the neighborhood radius. The learning rate controls the extent to which
neuron weights are updated, and it decreases over time. The neighborhood radius defines the
area of influence around the winning neuron during the cooperation phase, and it also
decreases over time.
5. Topology Preservation: One of the key features of SOM is its ability to preserve the
topological properties of the input data in the output space. This means that similar input
vectors will map to nearby neurons in the SOM grid.
6. Convergence: The training process continues for a specified number of iterations or until a
convergence criterion is met, which may include the SOM grid no longer changing
significantly.
4
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
7. Visualization and Clustering: Once trained, the SOM can be used for various purposes. It
can visualize high-dimensional data in a lower-dimensional grid, making it easier to
understand and interpret complex datasets. It can also be used for clustering, where similar
input vectors are grouped together on the SOM grid.
SOMs have found applications in various fields, including data mining, image analysis, and feature
extraction. They are particularly useful when dealing with high-dimensional data or when you want to
explore the structure of data in an unsupervised manner.
5
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Self-Organizing Map (SOM):
In the context of the paper, not only does it leverage the traditional Self Organizing Map (SOM)
algorithm, but it also takes a significant leap by introducing self-building versions of SOM. These novel
adaptations represent a substantial advancement in the field of machine learning.
What sets these self-building versions apart is their innate ability to adapt and evolve incrementally in
response to the specific characteristics of the data they encounter. This adaptability is a game-changer,
especially in the context of smart cities, where data is as dynamic and unpredictable as the urban
environments they serve.
These self-building versions of SOM possess a remarkable feature: they can automatically create and
modify AI structures. This means that when the system encounters new data patterns or alterations in the
existing ones, it doesn't require manual intervention. Instead, it self-adjusts, ensuring that the AI
framework remains in sync with the evolving landscape of a smart city.
In practice, this equips the AI system with the agility and responsiveness needed to handle the constant
flux of data in a dynamic urban setting. Whether it's the sudden surge in traffic, the fluctuations in energy
consumption, or the ever-changing trends in social behavior, these self-building versions ensure that the
AI adapts seamlessly and effectively.
By introducing these innovative self-building versions of SOM, the paper embraces the core principles of
adaptability and real-time responsiveness—attributes that are quintessential for the successful
implementation of AI in the context of smart cities. This technology not only improves the accuracy and
relevance of data analysis but also enhances the overall efficiency and effectiveness of decision-making
processes in these complex urban environments.
6
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Self-Building Versions of SOM:
The self-building (structure-adapting) versions of Self-Organizing Maps (SOM) represent a significant
advancement in machine learning, particularly in the context of smart cities. Here's a breakdown of
what this means:
• Traditional SOM: Self-Organizing Maps are a type of artificial neural network used for
unsupervised learning. They're great for visualizing and understanding complex data patterns.
Traditional SOMs are usually pre-defined structures with a fixed number of nodes (neurons)
arranged in a grid.
• Self-Building Versions: These are an extension of the traditional SOM. Instead of having a
fixed structure, self-building SOMs can dynamically create and modify their structure. They
start with a small structure and incrementally grow or shrink their neural network based on the
specific characteristics of the data they are processing.
• Adapting to Data Characteristics: The key idea is that the SOM "learns" about the data as it
processes it. If the data becomes more complex, the SOM can add more neurons to capture this
complexity. If the data becomes simpler, it can reduce its structure to be more efficient. This
adaptability is crucial in the dynamic and ever-changing environment of a smart city.
• Evolving Incrementally: Instead of a static, one-size-fits-all approach, these self-building
SOMs can evolve over time. As the data in a smart city changes (which it often does due to
various factors like weather, traffic, events, etc.), the algorithm can automatically adjust its
structure to better represent and understand the new data patterns.
• Adaptation for Smart Cities: Smart cities generate vast amounts of data from various sources,
including sensors, cameras, and IoT devices. This data is diverse and can change rapidly. Self-
building SOMs are particularly useful in this context because they can continuously adapt to
the unique data characteristics of a smart city, helping in tasks like traffic management, energy
optimization, and more.
In essence, these self-building SOMs are like self-adjusting, data-driven neural networks. They offer a
flexible and adaptive approach to processing and making sense of the data in a smart city, where
traditional, fixed neural network structures might struggle to keep up with the ever-changing
environment.
10
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Process-flow:
• Problem: This is the initial step, where you identify the challenges and objectives related to data
analytics in a smart city context.
• Data Collection: Data is gathered from various sources, including sensors, social media, and
CCTV data.
• Preprocessing: Data preprocessing involves cleaning, transforming, and making data ready for
analysis.
• Data Integration: The collected data from different sources is integrated into a cohesive dataset.
• Real-Time Data Processing: This step involves processing real-time data to keep up with the
dynamic nature of smart city environments.
• AI & Machine Learning: AI and machine learning models are applied to the data to extract
meaningful insights and patterns.
• Self-Building AI: Self-building AI models adapt and evolve autonomously, allowing for
seamless adjustments to changing data patterns.
• Adaptive Models: These models continually adapt to the evolving data, making them well-suited
for dynamic smart city environments.
• Data Analytics: Insights and recommendations are generated through data analytics, enhancing
decision-making.
• Optimized Decisions: These optimized decisions contribute to better resource allocation and
overall city management.
• Deploy in Smart City: The adaptive models and insights are implemented in the smart city
environment.
• Monitor & Feedback Loop: The system continuously monitors performance and collects
feedback to improve further.
18
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
APPLICATIONS:
The paper proposes a framework based on self-building artificial intelligence (AI) and
machine learning techniques for the smart city environment. Here are some potential applications
of this paper's research:
1. Traffic Management: With the ability to generate global scenarios from local information,
the proposed framework can support decision-making processes in traffic management. It can
help identify traffic patterns, congestion areas, and potential causes such as road works or
large trucks. This information can be used to optimize traffic flow, divert traffic during
emergencies, and plan road or infrastructure repairs based on real-time data.
2. Emergency Response: The self-building AI framework can assist in emergency response
situations by providing real-time information and support to emergency services. It can help
with police and security positioning , enabling authorities to make data-driven decisions
based on the analysis of various data sources such as traffic, pedestrian movements, and
public functions in the smart city environment.
3. Public Services Optimization: The proposed framework has practical implications for
optimizing public services in smart cities. For example, it can be used to analyze data related
to the spread of diseases or epidemics within neighborhoods. This information can aid in
resource allocation, surveillance, and containment strategies. Additionally, it can assist in
optimizing the delivery of essential services such as electricity, gas, and water based on
demand patterns and infrastructure conditions.
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
4. Infrastructure Planning: The self-building AI framework can contribute to infrastructure
planning in smart cities. By analyzing data from various sources, including transportation,
energy consumption, and population density, the framework can help identify areas where
infrastructure improvements are needed. This information can be used for efficient urban
planning, optimizing resource allocation, and enhancing the overall functionality of the city's
infrastructure.
Overall, the applications of this research paper's framework extend to various domains within
he smart city environment, including traffic management, emergency response, public
services
Optimization, and infrastructure planning. By leveraging self-building AI in these areas,
cities can enhance their efficiency, sustainability, and resilience.
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Overview and framework:
Growing Self-Organizing Map:
The paper utilizes the Self Organizing Map (SOM) algorithm, and its extended version called the
Growing Self Organizing Map (GSOM), in the proposed AI framework for smart cities.
The GSOM algorithm is an unsupervised machine learning technique that organizes high-dimensional
data into a lower-dimensional representation. In the context of the paper, the GSOM is employed to
create local feature maps at the on-site local processors in the second layer of the AI framework.
The GSOM starts with a small initial map and dynamically grows as new data is presented to it. It
consists of a grid of nodes, where each node represents a prototype or a cluster in the input data space.
During the training phase, the GSOM adapts its nodes to better represent the input data distribution.
This adaptation is achieved by adjusting the prototypes' weights based on the similarity between the
input data and the prototypes.
The GSOM also supports topological preservation, meaning that neighboring nodes on the map are
likely to represent similar data patterns. This feature allows for spatial representation of the data,
enabling visualization of similarities and relationships among the input data.
In the proposed framework, the GSOM is utilized in the local layer for creating local feature maps
based on the specific data characteristics of the smart city environment. These feature maps capture the
underlying patterns and relationships within the data, enabling data-driven insights and decision-
making at the local level.
Additionally, the paper discusses the application of Distributed GSOM (DGSOM), an enhanced variant
of the GSOM algorithm, in the third layer of the framework. DGSOM allows for parallel and
distributed processing of data, enabling the creation of a global GSOM. The global GSOM supports
global analytics, providing insights into global patterns and facilitating coordination between the local
processing units in the smart city.
Overall, the use of the GSOM algorithm, along with its extended version, in the proposed framework
helps in effectively organizing and analyzing the complex data in the smart city environment, leading
to improved decision-making and optimization of various smart city applications
15
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
16
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
AI Proposed Framework Machine Learning Techniques:
The paper proposes an AI framework for smart cities based on unsupervised machine learning techniques,
specifically the Self Organizing Map (SOM) and self-building versions of SOM. The framework aims to address
the challenges of isolated and siloes AI applications in smart cities by integrating information and enabling
structure adaptation.
The proposed framework consists of three layers: on-site, local, and global. The on-site layer involves data
capturing from devices like CCTV cameras or sensors, followed by initial data pre-processing. This layer generates
the on-site GSOM (Self Organizing Map) to represent the local data.
17
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
The local layer utilizes self-building versions of the SOM algorithm to adapt and evolve based on the specific
characteristics of the data. This allows for the automatic creation and modification of AI structures. The local
processors use the self-building AI algorithm to develop local feature maps, providing data-driven insights and
decision-making capabilities.
The global layer, implemented as a platform within a cloud environment, facilitates inter-layer communication and
information sharing. It enables the coordination and updating of local feature maps across different sites in the
smart city. This global layer helps in integrating information from various sources and supports comprehensive
analysis and decision-making.
The framework's objective is to enable efficient and effective AI applications in smart cities by utilizing self-
building AI structures, integrating local and global processing, and leveraging cloud computing capabilities. The
proposed framework has the potential to optimize various aspects of smart cities, such as traffic management,
emergency response, public services optimization, and infrastructure planning.
Future work includes conducting specialized analytics and implementing the framework on a cloud environment to
further simulate the complete functionality of the framework. The research is supported by the Data to Decisions
Cooperative Research Centre (D2D CRC) and La Trobe University Postgraduate Research Scholarship.
Overall, the framework presented in the paper offers a comprehensive solution for integrating and adapting AI
techniques in smart cities, paving the way for more efficient and intelligent urban environments.
18
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
LIMITATIONS:
While self-building AI and machine learning have the potential to bring significant improvements
to smart cities, they also come with limitations and challenges. Here are some of the key limitations in
the context of smart cities:
• Data Privacy and Security Concerns: Smart cities generate vast amounts of data, often
including sensitive information about citizens. The use of self-building AI and machine learning
can raise concerns about data privacy and security, leading to potential breaches or misuse of
data.
• Bias and Fairness Issues: AI and machine learning models can inherit biases present in the data
they are trained on. In the context of smart cities, this could lead to unfair or discriminatory
outcomes, particularly in areas like law enforcement, public services, and resource allocation.
• Data Quality and Reliability: The quality of data collected in smart cities can vary, and it may
contain errors or inconsistencies. Self-building AI and machine learning heavily rely on data
quality, and inaccurate or unreliable data can lead to incorrect decisions and actions.
• Interoperability Challenges: Smart cities often employ diverse technologies and systems.
Ensuring interoperability between different AI and machine learning applications can be complex
and require significant coordination and standardization efforts.
• Complexity and Cost: Implementing self-building AI and machine learning systems in smart
cities can be expensive and technically complex. This includes the cost of hardware, software,
data management, and ongoing maintenance.
• Lack of Transparency: Many AI and machine learning models are considered "black boxes,"
making it difficult to understand the reasoning behind their decisions. In the context of smart
cities, transparency is crucial for accountability and trust.
• Ethical Considerations: The deployment of AI in smart cities raises ethical questions about
surveillance, data collection, and the use of technology for various purposes, requiring careful
ethical considerations and guidelines.
• Resistance to Change: Citizens and stakeholders may resist the introduction of new technologies
and AI-driven systems due to concerns about job displacement, privacy, and the loss of human
decision-making.
19
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
• Vulnerability to Attacks: AI and machine learning systems can be vulnerable to adversarial
attacks, where malicious actors manipulate input data to cause misbehavior in the systems,
potentially leading to security breaches or unintended consequences.
• Resource Constraints: Many cities, especially smaller ones, may lack the necessary resources
and expertise to implement and maintain sophisticated AI and machine learning systems
effectively.
• Environmental Impact: The hardware and infrastructure required for AI and machine learning
can have a significant environmental impact due to energy consumption and electronic waste.
Addressing these limitations in the context of smart cities requires a comprehensive approach that
combines technological solutions, ethical considerations, policy frameworks, and ongoing community
engagement. It's essential to strike a balance between the benefits of self-building AI and machine
learning and the potential risks and challenges they pose to ensure the responsible and sustainable
development of smart cities.
1
10
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
Conclusion & Future Scope
Self-building AI and machine learning play a pivotal role in the context of smart cities.
These technologies offer numerous advantages, such as improving efficiency, enhancing
sustainability, and promoting a better quality of life for urban residents. Some key takeaways
include:
1. Efficient Resource Management: Self-building AI can help optimize resource allocation,
reducing waste and energy consumption.
2. Traffic Management: Machine learning can assist in real-time traffic management,
reducing congestion and improving transportation systems.
3. Sustainability: AI can facilitate the monitoring and control of energy use, leading to more
sustainable practices.
4. Public Services: Self-building AI can enhance the delivery of public services, such as
healthcare, security, and emergency response.
5. Data-Driven Decision-Making: Machine learning allows for data analysis and predictive
modeling, aiding city planners in making informed decisions.
6. Future Scope: The future of self-building AI and machine learning in smart cities is
promising and involves several areas of development:
7. Enhanced Automation: AI systems will become more autonomous in managing various
city functions, from waste management to public transport.
8. Predictive Maintenance: Machine learning will be used to predict infrastructure
maintenance needs, ensuring long-term sustainability.
1
11
Name: Zainab Sayyed Roll no.:13327, A.Y.2023-24
Sem-I
9. 5G and IoT Integration: The integration of 5G and IoT will enable faster data collection
and analysis, allowing for more real-time decision-making.
10. Urban Planning: AI will help in designing urban spaces for optimized resource usage,
improved traffic flow, and reduced environmental impact.
11. Citizen Engagement: AI-driven platforms will engage citizens in decision-making
processes, increasing their involvement in shaping the city.
12. Security: AI will enhance the security and resilience of smart city systems, safeguarding
against cyber threats and emergencies.
In conclusion, self-building AI and machine learning have the potential to revolutionize
how smart cities operate, making them more efficient, sustainable, and citizen-centric. As
technology continues to advance, it will be crucial for city planners and developers to embrace
these innovations to build smarter, more connected, and more livable cities.
1
12
References
1. Adikari A., De Silva, D., Alahakoon, D., & Yu, X. (2019). A cognitive model for emotion awareness in
industrial Chatbots. 2019 IEEE 17th international conference on industrial informatics (INDIN), 1, 183–
186.
2. Alahakoon, D., Halgamuge, S. K., & Srinivasan, B. (2000). Dynamic self-organizing maps with controlled
growth for knowledge discovery. IEEE Transactions on Neural Networks, 11(3), 601–614.
https://doi.org/10.1109/72.846732.
3. Allam, Z., & Dhunny, Z. A. (2019). On big data, artificial intelligence and smart cities. Cities, 89, 80–91.
Alter, S. (2019). Making sense of smartness in the context of smart devices and smart systems.
4. Information Systems Frontiers, 1–13. Bandaragoda, T., Adikari, A., Nawaratne, R., Nallaperuma, D.,
Luhach, A, Kr., Kempitiya, T., Nguyen, S., Alahakoon, D., De Silva, D., & Chilamkurti, N. (2020).
5. Artificial intelligence based commuter behavior profiling framework using internet of things for real-time
decision-making. Neural Computing and Applications. https://doi. org/10.1007/s00521-020-04736-7.
Boccaletti, S., Grebogi, C., Lai, Y. C., Mancini, H., Maza, D., & Lai, Y.- C. (2000).
6. The control of chaos: Theory and applications. Physics Reports, 329(3), 103–197.
https://doi.org/10.1016/S0370-1573(99) 00096-4. Bundy, A. (2017).
7. Preparing for the future of artificial intelligence. Springer Carey, L., Walsh, A., Adikari, A., Goodin, P.,
Alahakoon, D., De Silva, D., Ong, K.-L., Nilsson, M., & Boyd, L. (2019).
8. Finding the intersection of neuroplasticity, stroke recovery, and learning: Scope and contributions to stroke
rehabilitation. Neural Plasticity., 2019, 1–15. https://doi.org/10.1155/2019/5232374. Chen, Y., Qin, B.,
Liu, T., Liu, Y., & Li, S. (2010). The comparison of SOM and K-means for text clustering.
9. Computer and Information Science, 3(2), 268–274. Chen, B.-W., Imran, M., Nasser, N., & Shoaib, M.
(2019). Self-aware Autonomous City: From sensing to planning. IEEE Communications Magazine, 57(4),
33–39. https://doi.org/10.1109/ MCOM.2019.1800628. Choi, W., Shahid, K., & Savarese, S. (2009).
10. What are they doing?: Collective activity classification using spatio-temporal relationship among people.
Computer vision workshops (ICCV workshops), 2009 IEEE 12th international conference on, 1282–1289.
Cziko, G, A. (2016).
11. Unpredictability and indeterminism in human behavior: Arguments and implications for educational
Research: Educational Researcher. https://doi.org/10.3102/ 0013189X018003017, Dalal, N., & Triggs, B.
(2005).
12. Histograms of oriented gradients for human detection. 2005 IEEE Computer Society Conference on
Computer Vision and Pattern Recognition (CVPR’05), 1, 886–893 vol. 1.
https://doi.org/10.1109/CVPR.2005.177.
20
13. De Silva, D., Sierla, S., Alahakoon, D., Osipov, E., Yu, X., & Vyatkin, V. (2020). Toward intelligent
industrial informatics: A review of current developments and future directions of artificial intelligence in
industrial applications.
14. IEEE Industrial Electronics Magazine, 14(2), 57–72. https://doi.org/10.1109/MIE.2019.2952165.
15. Eldrandaly, K. A., Abdel-Basset, M., & Abdel-Fatah, L. (2019). PTZsurveillance coverage based on
artificial intelligence for smart cities.
16. International Journal of Information Management., 49, 520–532.
https://doi.org/10.1016/j.ijinfomgt.2019.04.017. Emam, A. (2015). Intelligent drowsy eye detection using
image mining. Information Systems Frontiers, 17(4), 947–960. Fonseka, A., & Alahakoon, D. (2010).
17. Exploratory data analysis with multi-layer growing self-organizing maps. 2010 fifth international
conference on information and automation for sustainability, 132– 137.
https://doi.org/10.1109/ICIAFS.2010.5715648. Inf Syst Front (2023) 25:221–240 237
18. Fonseka, A., Alahakoon, D., & Rajapakse, J. (2011). A dynamic unsupervised laterally connected neural
network architecture for integrative pattern discovery. In B.-L. Lu, L.
19. Zhang, & J. Kwok (Eds.), Neural information processing (pp. 761–770). Berlin Heidelberg: Springer.
Fritzke, B. (1994). Growing cell structures—A self-organizing network for unsupervised and supervised
learning. Neural Networks, 7(9), 1441–1460.
20. Ganegedara, H., & Alahakoon, D. (2011). Scalable data clustering: A Sammon’s projection based
technique for merging GSOMs. In B.- L. Lu, L. Zhang, & J. Kwok (Eds.), Neural information processing
(pp. 193–202).
21. Berlin Heidelberg: Springer. Ganegedara, H., & Alahakoon, D. (2012). Redundancy reduction in self-
organising map merging for scalable data clustering. The 2012 International Joint Conference on Neural
Networks (IJCNN), 1–8. https://doi.org/10.1109/IJCNN.2012.6252722.
22. Garcia-Font, V., Garrigues, C., & Rifà-Pous, H. (2016). A comparative study of anomaly detection
techniques for Smart City wireless sensor networks. Sensors (Basel, Switzerland), 16(6). https://doi.org/
10.3390/s
20
