Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Ahmed Aboulfotouh

Thesis title: User-Centric Clustering and Pilot Assignment in Cell-Free Networks


Abstract: Current 5G networks are primarily built on the cellular massive MIMO physical layer technology which achieved significant improvement in spectral efficiency as compared to previous generations. Nevertheless, there is always an increasing demand for higher data rates, and more reliable and uniform service. After successful massive MIMO deployments, it has become a natural question, "what will the physical layer in beyond 5G and 6G networks be like?"

Cell-free massive MIMO has emerged as a physical layer technology that is one of the candidates to takeover future deployments in beyond 5G and 6G networks. The main concept is to go beyond the cellular paradigm by employing an ultra dense deployment of small-sized multi-antenna access points (APs) which cooperate to serve users in the coverage area, eliminating the notion of boundaries between cells. The cell-free architecture has shown the capability of providing uniform service within the coverage area, while cellular networks suffer from poor performance at cell edges. It also has better ability to manage interference due to cooperation between APs which is not the case in cellular networks with no cooperation.

The most practical form of this paradigm is user-centric cell-free massive MIMO. Instead of allowing all the APs to serve all the users in the network, each user is served by a subset of the APs which ensures that network operation is scalable as the number of users grows. The main objective of this thesis is to provide a structured approach to design the cluster of APs that serve each user which is known as the user-centric clustering problem. In the pursuit to solve the clustering problem, there is another problem which is tightly connected to it, the pilot assignment problem. Both problems have to be solved together to ensure satisfactory network-wide performance.

The thesis provides a mathematical formulation for each of the user-centric clustering and the pilot assignment problems as stochastic non-linear binary integer programs which are solved using sample average approximation and the genetic algorithm. The pilot assignment problem is formulated such that it takes into account the usercentric clusters while choosing the pilot assignments which makes the optimization more accurate and efficient. Numerical experiments show that the resulting solutions outperform heuristic baseline algorithms from the literature, leading to significant spectral efficiency gains. Furthermore, we propose an approximate approach to derive closed form expressions of the uplink and downlink SINR which eliminates the need for sampling.

In future work, more effort shall be directed towards finding more practical approaches to realize the optimized solutions of both the user-centric clustering and pilot assignment problems with reasonable time-complexity. For instance, one candidate approach is to use machine learning and AI methods to learn structured approaches to solve both problems, using the optimized solutions as a reference for learning.


Committee Members:
Dr. Zubair Fadlullah (supervisor, committee chair), Dr. Muhammad Asaduzzaman, Dr. Al-Sakib Khan Pathan (United International University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Title: "Biosynthesis of ω–Hydroxy Fatty Acid Polyesters in Nicotiana tabacum Stigmas"

Supervisory Committee:
Dr. Wensheng Qin (Supervisor)
Dr. Isabel Molina (Co-Supervisor)
Dr. Heidi Schraft
Dr. Doug Reid
Dr. Guanqun(Gavin) Chen (External)

All are welcome to attend. Please contact biology@lakeheadu.ca for the meeting ID and password.

Lakehead University International and the Office of Indigenous Initiatives invite all students, staff and faculty to join us for the KAIROS Blanket Exercise.

The KAIROS Blanket Exercise (KBE) is an interactive learning experience that teaches the Indigenous rights history. Developed in response to the 1996 Report of the Royal Commission on Aboriginal Peoples—which recommended education on Canadian-Indigenous history as one of the key steps to reconciliation, the Blanket Exercise covers over 500 years of history in a one-and-a-half-hour participatory workshop.

Blanket Exercise participants take on the roles of Indigenous peoples in Canada. Standing on blankets that represent the land, they walk through pre-contact, treaty-making, colonization and resistance. They are directed by facilitators representing a narrator (or narrators) and the European colonizers.

Participants are drawn into the experience by reading scrolls and carrying cards which ultimately determine their outcomes. By engaging on an emotional and intellectual level, the Blanket Exercise effectively educates and increases empathy. The exercise will be followed by a debriefing session in which participants have the opportunity to discuss the experience as a group. This often takes the form of a talking circle.

Please register in advance at https://forms.gle/THKHU4XaZKVX7N4S9.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Khaled Bedda

Thesis title: New Paradigms of Distributed AI for Improving 5G Network Performance


Abstract: With the advent of 5G technology, there is an increasing need for efficient and effective machine learning techniques to support a wide range of applications, from smart cities to autonomous vehicles. The research question is whether distributed machine learning can provide a solution to the challenges of large-scale data processing, resource allocation, and privacy concerns in 5G networks. The thesis examines two main approaches to distributed machine learning: split learning and federated learning. Split learning enables the separation of model training and data storage between multiple devices, while federated learning allows for the training of a global model using decentralized data sources. The thesis investigates the performance of these approaches in terms of accuracy, communication overhead, and privacy preservation. The findings suggest that distributed machine learning can provide a viable solution to the challenges of 5G networks, with split learning and federated learning techniques showing promising results for spectral efficiency, resource allocation, and privacy preservation. The thesis concludes with a discussion of future research directions and potential applications of distributed machine learning in 5G networks.

In this thesis, we investigate the application of distributed machine learning in 5G networks, with a focus on split learning and federated learning techniques for improving spectral efficiency and resource allocation. The research question of whether distributed machine learning can provide comparable performance to centralized methodologies is explored using a novel method of split learning called peer-coordinated sequential split learning and asynchronous federated learning. The results show that the peer-coordinated sequential split learning method provides comparable performance to the vanilla method, while the asynchronous federated learning technique performs well in terms of low latency and network overhead. These findings suggest that distributed machine learning can be effectively applied in 5G applications, improving network performance in terms of spectral efficiency and resource allocation.

We propose two methodologies; PC-SSL peer-coordinated sequential slit learning and AFD Asynchronous federated learning. The proposed PC-SSL minimizes the data transmitted between the client BSs and a server by processing data locally on the clients. This results in low latency and computation overhead in making handoff decisions and other network operations. While the Federated learning model permits a reasonably accurate decision for the resource allocation for different 5G users without violating their privacy or introducing additional load to the network. Experimental results demonstrate the efficiency of the asynchronously weight updating federated learning in contrast with the conventional FedAvg (Federated averaging) strategy and the traditional centralized learning model. In particular, our proposed technique achieves network overhead reduction with a consistent and significantly high prediction accuracy, that validates its low-latency and efficiency advantages



Committee Members:
Dr. Zubair Fadlullah (supervisor, committee chair), Dr. Muhammad Asaduzzaman, Dr. Al-Sakib Khan Pathan (United International University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Anwar As'ad

Thesis title: Moreau Envelopes-based Personalized Asynchronous Federated Learning: Improving Practicality in Distributed Machine Learning


Abstract: Federated learning is a promising approach for training models on distributed data, driven by increasing demand in various industries. However, it faces several challenges, including communication bottlenecks and client data heterogeneity. Personalized asynchronous federated learning addresses these challenges by customizing the model for individual users based on their local data while trading model updates asynchronously. In this paper, we propose Personalized Moreau Envelopes-based Asynchronous Federated Learning (APFedMe) that combines personalized learning with asynchronous communication and Moreau Envelopes as clients’ regularized loss functions. Our approach uses the Moreau Envelopes to handle non-convex optimization problems and employs asynchronous updates to improve communication efficiency while mitigating heterogeneity data challenges through a personalized learning environment. We evaluate our approach on several datasets and compare it with PFedMe, FedAvg, and PFedAvg federated learning methods. Our experiments show that APFedMe outperforms other methods in terms of convergence speed and communication efficiency. Then, we mention some well-performing implementations to handle missing data in distributed learning. Overall, our work contributes to the development of more effective and efficient federated learning methods that can be applied in various real-world scenarios.



Committee Members:
Dr. Zubair Fadlullah (supervisor, committee chair), Dr. Muhammad Asaduzzaman, Dr. Al-Sakib Khan Pathan (United International University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Aya Farrag

Thesis title: Towards a Trustworthy End-to-End AI-based Clinical Decision Support System for Disease Diagnosis and Prognosis: Breast Cancer Use-Case


Abstract: Artificial Intelligence (AI) research has emerged as a powerful tool for health-related applications. With the increasing shortage of radiologists and oncologists around the world, developing an end-to-end AI-based Clinical Decision Support (CDS) system for fatal disease diagnosis and survivability prediction can have a significant impact on healthcare professionals as well as patients. Such a system uses machine learning algorithms to analyze medical images and clinical data to detect cancer, estimate its survivability and aid in treatment planning. We can break the CDS system down into two main components: the Computer Aided Diagnosis (CAD) subsystem and the Computer-Aided Prognosis subsystem (CAP). The lack of trustworthiness of these systems is still considered a challenge that needs to be addressed in order to increase their adoption and usefulness in real-world applications.

In this thesis, we propose new methods and frameworks to address existing challenges and research gaps in both components of the system to pave the way toward its usage in clinical practice. We select the breast cancer use case to perform our analysis as it is the most common cancer amongst women and the second leading cause of cancer death.

In cancer CAD systems, the first and most important step is to analyze medical images to identify potential tumors in a specific organ. In dense prediction problems like mass segmentation, preserving the input image resolution plays a crucial role in achieving good performance. However, this resolution is often reduced in current Convolution Neural Networks (CNN) that are commonly repurposed for this task. In Chapter 3, we propose a double-dilated convolution module in order to preserve spatial resolution while having a large receptive field. The proposed module is applied to the tumor segmentation task in breast cancer mammograms as a proof-of-concept. To address the pixel-level class imbalance problem in mammogram screenings, different loss functions (i.e., binary cross-entropy, weighted cross-entropy, dice loss, and Tversky loss) are evaluated. Experimental analysis is performed to compare the performance of lesion segmentation networks on mammogram screenings before and after plugging the proposed module into one state-of-the-art deep convolutional neural network. The obtained results show the effectiveness of the proposed module in increasing the similarity score and decreasing the miss-detection rate.

Following the cancer diagnosis step, in Chapter 4, we propose a new framework for cancer survival prediction in CAP systems to precisely predict the estimated survival months of patients in order to facilitate treatment planning. We combine two main strategies in solving the cancer survivability prediction problem using Machine Learning techniques. In the first strategy, we model the survivability prediction task as a two-step problem, namely a classification problem to predict whether or not a patient survives for five years, and a regression problem to forecast the number of remaining months for those who are predicted to not survive for five years. The second strategy is to develop stage-specific models, where each model is trained on instances belonging to a certain cancer stage in order to precisely predict survivability of patients from the same stage. We investigate the impact of adopting these strategies along with applying different balancing techniques over the model performance using breast cancer clinical data. The obtained results demonstrate that the proposed methods prove effective in both survivability classification and regression.

To incorporate the role of prognosis in determining the most suitable treatment plans for a cancer patient, in Chapter 5, we propose a novel prognostic-based framework for treatment planning. We employ the prediction models developed for stage-specific survival estimation to determine the best possible treatment plans in terms of prognostic outcomes. The system generates an ordered list of all possible combinations of treatments associated with their predicted survival outcomes to offer more comprehensive and intuitive treatment recommendations that can aid medical professionals in making more informed decisions about the most appropriate course of treatment for their patients. By integrating survival prediction models into treatment planning, healthcare providers can offer better patient care and help patients and their families make more informed decisions about their treatment options.

Experiments conducted in different chapters of this thesis demonstrate that the proposed AI-enabled techniques may potentially be utilized in future Clinical Decision Support Systems to help clinicians make patient-specific assessments and treatment decision



Committee Members:
Dr. Zubair Fadlullah (supervisor, committee chair), Dr. Muhammad Asaduzzaman, Dr. Al-Sakib Khan Pathan (United International University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Gad Gad

Thesis title: Light-weight Federated learning with Augmented Knowledge Distillation for Human Activity Recognition


Abstract: The field of deep learning has experienced significant growth in recent years in various domains where data can be collected and processed. However, as the data plays a central role in the deep learning revolution, there are risks associated with moving the data from where it is produced to central servers and data centers for processing.

To address this issue, Federated Learning (FL) was introduced as a framework for collaboratively training a global model on distributed data. However, FL comes with several challenges, including communication overhead, system and statistical heterogeneity, and privacy concerns. To address these challenges, this thesis proposes the incorporation of techniques such as Knowledge Distillation (KD) and Differential Privacy (DP) with FL. Specifically, a modelagnostic FL algorithm based on KD is proposed, called the Federated Learning algorithm based on Knowledge Distillation (FedAKD). FedAKD utilizes a shared dataset as a medium to calculate soft labels (knowledge), which are then sent to the server for aggregation and broadcast back to clients to train on them in addition to local training. Additionally, the privacy analyses of the uplink and the downlink channels of federated learning based on the differential privacy (DP) definition is discussed, and the performance of FedAKD is evaluated using the human activity recognition (HAR) application with four datasets.

• In chapter 2, various concepts that are employed in the following chapters of this thesis are covered, including human activity recognition, deep learning models, federated learning, and empirical loss minimization (ELM).
• Chapter 3 introduces FedAKD, a communication-efficient federated learning method, and evaluates its performance using the human activity recognition (HAR) application with four datasets.
• Chapter 4 discusses the communication efficiency challenge of deploying federated learning on IoT networks and evaluates the communication overhead of FedAKD.
• Chapter 5 focuses on the privacy analysis of federated learning by presenting a realistic federated learning threat model and providing privacy guarantees based on differential privacy (DP).

The experiments show that FedAKD achieves better performance than other KD-based FL algorithms and comparable performance to model-based FL methods. Furthermore, decreasing the privacy budget causes a slight degradation in performance, as DP clips the gradients during training and adds noise to them to limit per-sample contribution to model weights, thus protecting privacy.

In conclusion, communication-efficient FL algorithms with privacy-preserving techniques present a viable solution for distributed learning while tolerating system heterogeneity and providing provable privacy guarantees, particularly in IoT applications.


Committee Members:
Dr. Zubair Fadlullah (supervisor, committee chair), Dr. Muhammad Asaduzzaman, Dr. Al-Sakib Khan Pathan (United International University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Lakshmi Preethi Kamak

Thesis title: Programming Pedagogy in the Age of Accessible Artificial Intelligence


Abstract: Human-Computer Interaction and Human-Artificial Intelligence Interaction research combined with Programming pedagogy research is evolving modern classrooms in many aspects with the aim of an optimal user experience. Incorporating these Human-Computer Interaction research principles can help novice programmers overcome cognitive overload and poor user experience. This thesis proposes methods to cater to students’ different learning styles and needs to enhance programming pedagogies in various learning environments. Providing a foundation for this thesis in terms of research direction and trending techniques, Chapter 2 describes the increasing popularity of and convergence of CT and AI research in programming pedagogy. Chapter 3 offers an innovative approach to teaching programming to middle school students using scaffolding techniques and computational thinking concepts in a blended learning environment for remote schools. More specifically, this study lays out a comprehensive curriculum and assessment materials aligned with Ontario learning outcomes with the mission to remove geographical barriers to building sustainability in First Nation Schools in Northwestern communities of Canada. Using an adapted survey instrument, we identified computationally talented students with the prospect of developing STEM interests. Similarly, Human-Artificial intelligence Interaction research principles can facilitate the integration of artificial intelligence tools to enhance programming understanding. Chapter 4 explores the efficacy of OpenAI’s ChatGPT as an artificial intelligence coding assistant in graduate students’ self-directed learning experience. In the study, students solved complex reinforcement learning tasks after completing a related Massive open online course. We developed an original survey instrument based on combining Human-Computer Interaction frameworks and Human-Artificial Intelligence Interaction metrics. It measures technology acceptance while verifying potential information bias based on an in-depth analysis of participant interaction behaviour. By leveraging these innovative techniques and technology and identifying potential limitations, this thesis seeks to enhance programming pedagogy, provide effective strategies for facilitating learning in diverse settings, and encourage future developments in this field.



Committee Members:
Dr. Vijay Mago (supervisor, committee chair), Dr. Jinan Fiaidhi, Dr. Piper Jackson (Thompson Rivers University)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.

Lakehead University Student Juried Exhibition and Honours Graduating Show will open to the public on the afternoon of Friday March 24th at Thunder Bay Art Gallery.

The Lakehead Department of Visual Arts invites you to join us for the opening and award ceremonies on Friday March 31, 2023 in Thunder Bay Art Gallery. Doors open at 7 pm. There will be food and refreshments at the opening.

This year the opening ceremonies will be emceed by Dr. Elizabeth Birmingham dean of Faculty of Social Sciences and Humanities.

Lakehead student exhibits will be on display until Sunday April 23rd.

Please join the Computer Science Department for the upcoming thesis defense:

Presenter: Archana Mariappan

Thesis title: Diagnosis and Level-I Therapeutical Support of Depression in Mental Health using Conversational AI


Abstract: World Health Organization statistics indicate that one out of every eight people suffers from mental illness. Due to the fear of stigma and social discrimination, they start being resilient and end up going through difficult situations alone. They fear criticism and start isolating them from friends, family and neighbours. They fear that may be called as ‘crazy’ which may affect their education, career and daily lives in a negative manner. Mostly people aren’t much aware about the mental health illness which may lead them into complicated consequences. Also, most of the people lack the knowledge about the mental health issues and trying to solve it alone without knowing the complexity of the future problems related to it. Also, people find it difficult to open up about their personal information and feelings to a stranger or counsellor assuming that their data wouldn’t be confidential. The majority of individuals doesn’t have access to effective care. If the issue isn’t treated with care it can lead to serious mental problems such as it may cause depression, obsessive compulsive disorder, anxious or personality disorder. Literally, majority of the population have inadequate knowledge about the mental health illness. In order to overcome this problem, our mental health chatbot was created. Our study aims to provide efficient and essential care to the people with mental health concerns according to their needs and supplying the basic information regarding mental health problems through various sources. The proposed system eases the diagnosis of mental health problem in the user by identifying and providing level-I therapeutical support for depression by employing conversational AI. This paper utilizes technologies like Artificial Intelligence and its subfield Natural Language Processing (NLP) to provide a amicable environment for the user 24/7 and it can be integrated in cross platforms like iOS, android and windows etc. This system allows the users to schedule appointment, to learn in detail about the terminologies of mental health and it provides resources for feel good activities like videos and music. Through this system they can candidly express their feelings to the conversational AI chatbot besides their insecurity. The Artificial Intelligence (AI) in turn provides them with chat support, acts as a bridge to understand the situations and suggests solutions depending on the level of mental health deterioration. We propose a fully automated and powerful first-level detection and support system for mental health.


Committee Members:
Dr. Jinan Fiaidhi (supervisor, committee chair), Dr. Sabah Mohammed, Dr. Carlos Zerpa (Kinesiology)

Please contact grad.compsci@lakeheadu.ca for the Zoom link.
Everyone is welcome.