FUNDAMENTALS OF NEURAL NETWORKS
- Course
- AIEN421 - FUNDAMENTALS OF NEURAL NETWORKS
- Department
- Artificial Intelligence Engineering - English - Undergraduate
- Course Type
- Course
- Status
- Required
- Language
- English
- Credit
- 3
- ECTS
- 0
- T+P+L
- 3 + 0 + 0
- Course Coordinator(s)
- Asst. Prof. Dr. Tolga YIRTICI
- Prerequisite
- -
- Keywords
Course Description
The neural networks course represents the basic neural network architectures and learning algorithms which are inspired from the human brain. Neurocomputing stands as a sub-topic of artificial intelligence together with evolutionary computing. Within the scope of the course, perceptrons and perceptron learning algorithm are introduced. The architectures include Multilayer Perceptrons, Hopfield Model and Self Organized Maps covering Mallberg’s and Kohonen’s models. Learning algorithms will cover the selected supervised and unsupervised learning algorithms including backpropagation, Hebbian learning, competitive learning and the selected clustering algorithms. Several pattern recognition and computer vision applications will be presented as case studies. The applications will be simulated on selected simulators.
FUNDAMENTALS OF NEURAL NETWORKS
Evaluation Tools (Active Term)
| Item | Type | Weight (%) |
|---|---|---|
| Midterm | Midterm | 35 |
| Final | Final | 45 |
| Quiz 1 | Quiz | 10 |
| Quiz 2 | Quiz | 10 |
| Total | 100 | |
Course outcomes
No course outcomes have been defined yet.
Course Syllabus
| Week | Topic |
|---|---|
| Week 1 | Introduction to Neural Networks (Biological vs. artificial neurons, history, ML paradigms, applications) |
| Week 2 | Feedforward Neural Networks (Perceptron, MLP, forward propagation, activation functions) |
| Week 3 | Feedforward Neural Networks (Perceptron, MLP, forward propagation, activation functions) |
| Week 4 | Training Neural Networks (Loss functions, gradient descent, backpropagation) |
| Week 5 | Improving Neural Networks (Learning rate, momentum, overfitting, L1/L2, dropout, batch normalization, label smoothing) |
| Week 6 | Radial Basis Function Networks (Gaussian basis functions, K-means, center selection, pseudo-inverse, gradient descent) |
| Week 7 | QUIZ 1 && Self-Organizing Maps (SOM) (Competitive learning, BMU, neighborhood, clustering, visualization) |
| Week 8 | Midterm Week |
| Week 9 | Midterm Week |
| Week 10 | Grow-When-Required (GWR) & ECoS (Incremental learning, adaptive growth, online learning) |
| Week 11 | Hopfield Networks (Associative memory, Hebbian learning, energy function, convergence) |
| Week 12 | Time-Delay & Elman Networks (Tapped delay line, teacher forcing, recurrent connections) |
| Week 13 | GRU & LSTM (Vanishing gradients, gates, sequence learning) |
| Week 14 | Convolutional Neural Networks (Convolution, filters, padding, stride, pooling, CNN architectures) |
| Week 15 | QUIZ 2 && Model Evaluation & Modern Applications (Train/validation/test split, confusion matrix, precision, recall, F1, ethics, applications) |
Reference Books & Course Materials
- 01 Haykin, S. Neural Networks and Learning Machines, 3rd Edition, Pearson, 2009.
- 02 Aggarwal, C. C. Neural Networks and Deep Learning: A Textbook, Springer, 2018.
- 03 Goodfellow, I., Bengio, Y., & Courville, A. Deep Learning, MIT Press, 2016.
- 04 Géron, A. Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow, 3rd Edition, O'Reilly Media, 2022.
Learning Outcomes
No learning outcomes have been defined.
Program Outcomes
- P01 Apply data science principles and techniques to challenges in real life situations, and effectively communicate their solutions.
- P02 Identify and implement data analysis methodologies based on theoretical ideas, ethical code, and in-depth knowledge of the underlying data.
- P03 Analyze the guiding concepts and assessment procedures for information analysis in real-life applications.
- P04 Design and apply relevant data analysis models to find obscure solutions to business-related problems.
- P05 Utilize modern computing techniques to handle real-world problems characterized by massive amounts of data, such as parallel and distributed computing and machine learning.
- P06 Configure and administer the software tools required to efficiently produce usable information from any size of structured and unstructured datasets.
- P07 Administer or manage data science tools and techniques to organize and complete projects aimed at gaining useful insight from complex data.
- P08 Think critically and imaginatively, conceiving real-world issues from several angles, and work well in a variety of teams to solve issues cooperatively.
- P09 Be able to effectively integrate data‐based solutions into the user environment and help non-technical professionals in exploring, visualizing, and using these solutions
- P10 Understand their obligations under professional and ethical standards in relation to matters like data ownership and citation, data security and sensitivity and the privacy implications of data analysis.
Po-Lo Matrix
The PO-LO matrix has not been populated yet.