SECTION I: GENERAL INFORMATION ABOUT THE COURSE

Course Code	Course Name	Year	Semester	Theoretical	Practical	Credit	ECTS
60610METOS-YZM0389	Deep Learning	4	Spring	2	2	3	5

Course Type :	Elective Course IV
Cycle:	Bachelor      TQF-HE:6. Master`s Degree      QF-EHEA:First Cycle      EQF-LLL:6. Master`s Degree
Language of Instruction:	Turkish
Prerequisities and Co-requisities:	N/A
Mode of Delivery:	Face to face
Name of Coordinator:	Dr. Öğr. Üyesi ENVER AKBACAK
Dersin Öğretim Eleman(lar)ı:
Dersin Kategorisi:

SECTION II: INTRODUCTION TO THE COURSE

Course Objectives & Content

Course Objectives:

1. To familiarize students with Deep Learning Fundamentals. 2. To Introduce concepts for image, video and text recognition. 3. To teach students featured deep learning practices by a project. Textbook: http://neuralnetworksanddeeplearning.com/ and Deep Learning with Python (François Chollet) Software Tensorflow’s Keras Tools: Google’s Colab A long-term project will be assigned to students. The project reports will be prepared on any word processor in the following format. Midterm Exam 20% Final Exam 50% Projects 30%

Course Content:

Fundamentals What is deep learning, layers, learning, wieghts, loss functions, optimizers. Neural Networks Perceptrons, sigmoid neurons, learning with gradient descent and momentum based gradient descent. Neural Networks Back propagation. Cost Functions, cross entropy, softmax, regularization, wieght initialization, early stopping, batch size, tanh activation, dropout, normalization, relu layers, pooling layers. Mathematical fundamentals of Neural Networks Getting Started with Neural Networks Fundamentals of Deep Learning Supervised, unsupervised, selfsupervised and reinforcement learnings, train/validation/test splits allocations, vanishing gradient descent problem, overfitting and underfitting. Convolutional Neural Networks Fundamentals Pretrained models Fine-tuning Convolutional Neural Networks Feature extraction Python generators Visualizing intermediate activations and filters Class activation maps Encoder – Decoder Models Recurrent Neural Networks LSTMs, Combining CNN-LSTM models Video Processing 3D CNN, Keras functional API, callbacks Project Presentations

Course Specific Rules

Projects A long-term project will be assigned to students. The project reports will be prepared on any word processor in the following format. Page 1. • Project title • Student's name • Date due • Abstract (min 1/2 page) Page 2. Technical presentation. This section should include implemented technics and equations. (One page) Page 3 Discussion of findings. Significant findings in terms of the project objectives and reference any images generated. (Min two pages) Results. Includes all the images generated in the project. Number images individually so they can be referenced in the preceding section. Appendix. Computer codes with explanations written by the student. Projects not conform to the requested format may be grounds for rejection.

Course Learning Outcomes (CLOs)

Course Learning Outcomes (CLOs) are those describing the knowledge, skills and competencies that students are expected to achieve upon successful completion of the course. In this context, Course Learning Outcomes defined for this course unit are as follows:

Knowledge (Described as Theoritical and/or Factual Knowledge.)

Skills (Describe as Cognitive and/or Practical Skills.)

Competences (Described as "Ability of the learner to apply knowledge and skills autonomously with responsibility", "Learning to learn"," Communication and social" and "Field specific" competences.)

Weekly Course Schedule

Week	Subject	Materials Sharing *
		Related Preparation	Further Study
1)	Fundamentals What is deep learning, layers, learning, wieghts, loss functions, optimizers.
2)	Neural Networks Perceptrons, sigmoid neurons, learning with gradient descent and momentum based gradient descent.
3)	Neural Networks Back propagation.
4)	Neural Networks Cost Functions, cross entropy, softmax, regularization, wieght initialization, early stopping, batch size, tanh activation, dropout, normalization, relu layers, pooling layers.
5)	Mathematical fundamentals of Neural Networks
6)	Getting Started with Neural Networks
7)	Fundamentals of Deep Learning Supervised, unsupervised, selfsupervised and reinforcement learnings, train/validation/test splits allocations, vanishing gradient descent problem, overfitting and underfitting.
8)	Convolutional Neural Networks Fundamentals Pretrained models Fine-tuning
9)	Convolutional Neural Networks Feature extraction Python generators Visualizing intermediate activations and filters Class activation maps
10)	Encoder – Decoder Models
11)	Recurrent Neural Networks LSTMs, Combining CNN-LSTM models
12)	Video Processing 3D CNN, Keras functional API, callbacks
13)	Project Presentations
14)	Project Presentations

*These fields provides students with course materials for their pre- and further study before and after the course delivered.

Recommended or Required Reading & Other Learning Resources/Tools

Course Notes / Textbooks:

Hafta Konular Ders Kitabı Değerlendirme ve İlave Bilgiler 1 Derin Öğrenmeye Giriş Derin öğrenme temel kavramlar, Gerekli Matematiksel Tanımlamalar, Uygulamalarda kullanılacak arayüz ve yazılımların tanıtım ve kurulum uygulamalarını içermektedir. A 2 Yapay Sinir Ağları Temel Kavramlar Logistik Regresyon(LR), LR Cost Fonksiyonu, LR Gradient Descent, LR türev (derivatives) kavramlarını ve uygulamalarını içermektedir. A 3 Yapay Sinir Ağları (ANN) ANN aktivasyon fonksiyonları ve türev(derivative) hesaplamalarını ve ANN derin ağların oluşturulması içermektedir. A Projelerin Atanması 4 Yapay Sinir Ağları(ANN) ANN İleri-Geri Yayılım (Forward-Backward Propagation) tanıtım ve uygulamalarını içermektedir. A-B 5 Derin Öğrenme Stratejileri Derin Öğrenmede Bias Variance Kavramlarını, Performans Metriklerini ve Veri setlerinin Eğitim-Validasyon-Test Kümelerine ayrılma stratejilerini içermektedir. A-B 6 Derin Öğrenme Optimizasyon Parameterleri Regularization, Dropout, Normalization, Vanishing Gradient kavramlarını ayrıca Gradient Descent, Adam ve Rmsprop Optimizasyon Algoritmaları ile Learning Rate tanımlamaları ile Uygulamalarını içermektedir. A-B 7 Evrişimsel Sinir Ağları(CNN) Temelleri CNN matematiksel tanımlamalarını, Padding, Stride tanımlamaları ile çıkış haritalarının hesaplamalarını ayrıca temel CNN yapısı ve uygulamalarını içermektedir. B İlk Proje İncelemeleri 8 VIZE SINAVI 9 Evrişimsel Sinir Ağları(CNN) Modelleri-1 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) ‘de başarı sağlayan Alexnet VGG-16, VGG-19 model yapılarının incelenmesini ve Uygulamalarını içermektedir. B 10 Evrişimsel Sinir Ağları(CNN) Modelleri-2 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) ‘de başarı sağlayan Resnet, Inception Networks, Mobilenet model yapılarının incelenmesini ve Uygulamalarını içermektedir. B 11 Encoder – Decoder Modeller Autoencoder- Decoder modeller ve U-NET yapılarının incelenmesini ve Uygulamalarını içermektedir. PROJE SUNUMLARI-1 B İkinci Proje İncelemeleri 12 Tekrarlayan Sinir Ağları (RNN) Long Short-Term Memory (LSTM) Kavramları, CNN-LSTM Uygulamalarını içermektedir. PROJE SUNUMLARI-2 A-B İkinci Proje İncelemeleri 13 Evrişimli Sinir Ağları ile Nesne Tespiti YOLO ile nesne tespiti için tanımlamalar ve uygulamalarını içermektedir. PROJE SUNUMLARI-3 A-B İkinci Proje İncelemeleri 14 Final Sınavı

References:

Textbook: http://neuralnetworksanddeeplearning.com/ Deep Learning with Python (François Chollet) Software Tensorflow’s Keras Tools: Google’s Colab

Level of Contribution of the Course to PLOs

No Effect	1 Lowest	2 Low	3 Average	4 High	5 Highest

	Programme Learning Outcomes	Contribution Level (from 1 to 5)
1)	Has sufficient knowledge in mathematics, science, computer science and software engineering; use theoretical and applied knowledge in these fields together to solve software engineering problems
2)	Uses and applies theoretical and applied sciences in the field of basic science subjects for the solution of software engineering problems.
3)	Analyzes software engineering applications, designs and develops models to meet specific requirements under realistic constraints and conditions. For this purpose, selects and uses appropriate methods, tools and technologies.
4)	Identify, define, formulate and solve complex Software Engineering problems; for this purpose select and apply appropriate analytical and modeling methods
5)	Selects and effectively uses modern techniques and tools and information technologies required for computer science and Software Engineering applications.
6)	Designs a complex computer and software based system, process, device or product to meet certain requirements under realistic constraints and conditions, including economics, environmental issues, sustainability, manufacturability, ethics, health, safety, social and political issues; For this purpose, it applies modern design methods.
7)	Evaluates the norms and standards present in the works in which s/he takes responsibility in a critical point of view.
8)	Have the competencies required by the constantly developing field of Software Engineering and the global competitive environment.
9)	Communicates effectively in Turkish, both orally and in writing, has at least one foreign language knowledge at the level of European Language Portfolio B2

SECTION IV: TEACHING-LEARNING & ASSESMENT-EVALUATION METHODS OF THE COURSE

Teaching & Learning Methods of the Course

(All teaching and learning methods used at the university are managed systematically. Upon proposals of the programme units, they are assessed by the relevant academic boards and, if found appropriate, they are included among the university list. Programmes, then, choose the appropriate methods in line with their programme design from this list. Likewise, appropriate methods to be used for the course units can be chosen among those defined for the programme.)

Teaching and Learning Methods defined at the Programme Level	Teaching and Learning Methods Defined for the Course

Assessment & Evaluation Methods of the Course

(All assessment and evaluation methods used at the university are managed systematically. Upon proposals of the programme units, they are assessed by the relevant academic boards and, if found appropriate, they are included among the university list. Programmes, then, choose the appropriate methods in line with their programme design from this list. Likewise, appropriate methods to be used for the course units can be chosen among those defined for the programme.)

Aassessment and evaluation Methods defined at the Programme Level	Assessment and Evaluation Methods defined for the Course

Contribution of Assesment & Evalution Activities to Final Grade of the Course

Measurement and Evaluation Methods	# of practice per semester	Level of Contribution
Project	2	% 30.00
Midterms	1	% 20.00
Semester Final Exam	1	% 50.00
Total		% 100
PERCENTAGE OF SEMESTER WORK		% 50
PERCENTAGE OF FINAL WORK		% 50
Total		% 100

SECTION V: WORKLOAD & ECTS CREDITS ALLOCATED FOR THE COURSE

WORKLOAD OF TEACHING & LEARNING ACTIVITIES
Teaching & Learning Activities	# of Activities per semester	Duration (hour)	Total Workload
Course	0	0	0
Laboratory	0	0	0
Application	0	0	0
Special Course Internship (Work Placement)	0	0	0
Field Work	0	0	0
Study Hours Out of Class	0	0	0
Presentations / Seminar	0	0	0
Project	0	0	0
Homework Assignments	0	0	0
Total Workload of Teaching & Learning Activities	-	-	0
WORKLOAD OF ASSESMENT & EVALUATION ACTIVITIES
Assesment & Evaluation Activities	# of Activities per semester	Duration (hour)	Total Workload
Quizzes	0	0	0
Midterms	0	0	0
Semester Final Exam	0	0	0
Total Workload of Assesment & Evaluation Activities	-	-	0
TOTAL WORKLOAD (Teaching & Learning + Assesment & Evaluation Activities)			0
ECTS CREDITS OF THE COURSE (Total Workload/25.5 h)			5