Introduction to Machine Learning

LEARNING OUTCOMES

Basic purpose of this course is the familiarization of students with the topic of machine learning, namely with the processes involving a computer learning concepts without the need for direct programming/coding. It is a popular field of artificial intelligence (ΑΙ), which already finds a considerable range of applications (e.g. computer vision, speech identification/understanding, effective web search, medicine, autonomous driving). Successful completion of this course means that students:

• Have understood and can describe, analyze and compare the different categories of machine learning methods (supervised, unsupervised and reinforcement).
• Are able to apply and program optimization algorithms in application examples such as price prediction as well as classification of data in classes (linear and logistic regression).
• Have understood and are in position to select suitable forms of hypothesis functions (linear, non-linear) cost functions in regression algorithms (linear and logistic), and at the same time to apply tools for avoiding overparametarization.
• Understand and can apply algorithms of binary classification or of classification in more classes.
• Are able to check and evaluate the contribution of different input variables but also to split the training data into different groups for training and validation of machine learning algorithms. Τhey are also able to compute evaluation measures for different learning algorithms (precision, recall, f1).
• Have comprehended the operation principle of artificial neural networks and are able to train simple architectures in classification examples.
• Have understood and are in position to program clustering algorithms and apply them to image segmentation tasks.
• Have understood and are in position to apply methods and techniques for anomaly detection.
• Are able to reduce the number of variables of a machine learning algorithm by detecting via Principal Component Analysis which of these are uncorrelated.

 

General Competences

• Search for, analysis and synthesis of data and information, with the use of the necessary technology
• Working independently
• Working in an interdisciplinary environment
• Production of free, creative and inductive thinking

 

SYLLABUS

1. Introduction, historic review
2. Optimization methods (linear and logistic regression)
3. static/dynamic regression
4. regression with one and more variables
5. Supervised, unsupervised and reinforcement learning
6. Normalization
7. Artificial neural networks (models and architectures, forward-backward propagation)
8. Support Vector Machines (linear and non-linear classification)
9. Clustering (k-means, DBSCAN, Gaussian)
10. Dimensionality reduction (Principal Components Analysis)
11. Application examples and development of machine learning algorithms.

 

STUDENT PERFORMANCE EVALUATION

Language of evaluation: Greek
Methods of Evaluation:
• Written examination in the end of the semester (70%), which combines open-ended questions and numeric calculations.
• Evaluation of performance in the exercises (30%)

 

SUGGESTED BIBLIOGRAPHY

1. Bishop C., 2006. Pattern Recognition and Machine Learning. Springer-Verlag New York
2. Goodfellow I., Bengio Y., Courville A., Deep Learning. MIT Press

In Greek:
1. Simon Η., 2010. Neural Networks and Machine Learning. Papasotiriou Editions, Athens.
2. Diamantaras K., 2007. Artificial Neural Networks. Kleidaithmos Editions, Athens.