On Tuesday, 15th April 2025, the IEEE Communications Society BRAC University Student Branch Chapter hosted an engaging seminar titled “Mind the Network: How Neurons Connect the World” on campus. This proactive event delved into the parallels between neural structures in the human brain and artificial neural networks, exploring how these systems are reshaping the future of communication technologies. With a focus on both biological inspiration and machine learning architecture, the session offered a foundational understanding of the neural mechanisms behind smart technologies and their application in network systems. The seminar attracted an impressive audience of over 150 participants, including 35 IEEE members and a wide range of students from different departments. The interactive discussion, accompanied by real-life demonstrations and use cases, sparked curiosity and meaningful conversations, bridging the gap between neuroscience and network engineering. The event stood out as a collaborative step toward understanding how complex systems, biological and artificial, learn, connect, and develop in today’s digital age.
The event started on time and unfolded seamlessly. The host began by promoting the IEEE Week and the ongoing Membership Campaign, encouraging students to become part of the IEEE BRAC University Student Branch and its various chapters. Attendees were introduced to the week-long celebration, with special highlights on the events organized by IEEE BRACU SBC and all other affiliated chapters. Everyone was warmly invited to take part in this vibrant initiative.
Earlier in the morning, the inauguration ceremony of IEEE Week 2025 and the first-ever official merchandise launch of the IEEE BRAC University Student Branch took place. The ceremony was graced by the Honorable Registrar of BRAC University, along with the chapter advisors from all IEEE student chapters, making the moment truly prestigious. Following this joyous beginning, the seminar was held in the afternoon, serving as the very first event of IEEE Week and marking a memorable start to the week-long series of programs.
The esteemed speaker of this seminar was Moin Mostakim, Senior Lecturer, Department of Computer Science and Engineering, BRAC University.
Session by Moin Mostakim :
The session led by Moin Mostakim offered a comprehensive introduction to neural networks, emphasizing their foundational principles, internal mechanisms, and diverse applications. The session was divided into five focused segments, each designed to guide the audience through a progressive understanding of neural networks in both theory and practice.
The session began with the “Foundations of Neural Networks,” where Moin Mostakim drew inspiration from the human brain, highlighting its 86 billion neurons and intricate synaptic connections. This served as the conceptual bridge to artificial neural networks (ANNs), where input, hidden, and output layers work together to perform brain-like pattern recognition. He presented a comparative analysis of biological and artificial neurons, grounding the audience in how machine learning models are inspired by nature.
Next, in the “Core Components & Learning Process” segment, he delved into the roles of neurons, weights, biases, and activation functions. He clearly explained forward propagation—how data flows through layers—and backpropagation, the core mechanism behind learning. A focused discussion on gradient descent and its popular optimizers, such as SGD, Adam, and RMSprop, demonstrated the technical depth of neural network training.
The third segment explored “Neural Network Architectures,” providing an overview of ANN/MLPs, Convolutional Neural Networks (CNNs), and Recurrent Neural Networks (RNNs), including Long Short-Term Memory (LSTM) models. He concluded this part with a brief introduction to the transformative capabilities of transformer architectures, marking their growing relevance in natural language processing and large-scale tasks.
In “Optimization & Hyperparameter Tuning,” he outlined practical strategies for improving model performance, focusing on loss functions, learning rate, batch size, number of epochs, and architectural depth. Common issues like overfitting, underfitting, and the vanishing gradient problem were discussed, along with tips on designing and fine-tuning neural networks effectively.
The final segment, “Applications in Specialized Fields,” highlighted the real-world impact of neural networks. From reinforcement learning strategies like DQNs and policy gradients to the emerging field of Quantum Neural Networks (QNNs), the session shed light on cutting-edge research. A notable mention was made of large language models and their evolution through pretraining and fine-tuning techniques.
The session concluded with an interactive Q&A, where participants explored career pathways and clarified technical queries. Moin Mostakim’s clear, structured delivery and deep subject knowledge enabled attendees to grasp complex concepts with clarity and curiosity.
This session not only built a strong foundation in neural networks but also connected participants to their applications in modern communication systems, reinforcing the seminar’s central theme: “Mind the Network: How Neurons Connect the World.”
As the session came to a close, the advisor of IEEE Communications Society BRACU SBC, Tasfin Mahmud, took the stage to express gratitude and appreciation for the insightful presentation. To honor the distinguished speaker, a commemorative crest was presented as a gesture of appreciation for his valuable contributions to the seminar. Following this, a photo session was held with the speaker, along with the executive members of the chapter, capturing a moment of collective achievement and collaboration.
The seminar also ensured that all participants and guests were well taken care of, with refreshments provided. In recognition of the efforts and dedication of the volunteers, certificates of appreciation were presented as a token of gratitude for their invaluable support throughout the event.
With these final acknowledgments, the seminar successfully concluded, leaving participants with a deeper understanding of neural networks and their applications.