Computer Training Methods and Tools

Major: System Design
Code of subject: 7.122.03.O.002
Credits: 5.00
Department: Computer-Aided Design
Lecturer: Associate Professor at CAD department, PhD, Nazariy Andrushchak
Semester: 1 семестр
Mode of study: денна
Мета вивчення дисципліни: The purpose of studying the discipline "Methods and tools of machine learning" is to master the theoretical foundations of the main methods and tools used in machine learning, to develop students' intuition for a better understanding of the main ideas underlying these methods and tools, to teach students how to work with software that implements machine learning algorithms.
Завдання: The study of an academic discipline involves the formation and development of students' competencies: General competences: analytical thinking; solving non-standard problems; application of theoretical skills to solve applied problems. Professional competences: development of computer learning algorithms for appropriate application; the ability to program in the Python programming language.
Learning outcomes: As a result of studying the academic discipline, the student must be able to demonstrate the following learning outcomes: 1. implement and adjust computer training methods; 2. to check the statistical hypotheses of computer training; 3. solve regression recovery problems; 4. build decision trees and interpret the results of their operation for given models; 5. use in practice the method of support vectors for computer training; 6. design search engine algorithms; 7. analyze and develop algorithms for pattern recognition using computer learning methods; 8. evaluate the quality of computer learning algorithms. As a result of studying the academic discipline, the student must be able to demonstrate the following program learning outcomes: ЗН1 - The ability to formulate and improve an important research problem, to collect the necessary information for its solution, and to formulate conclusions that can be defended in a scientific context. ЗН7 - The ability to work with expert sources of information for the integration of data and knowledge in the field of the organization using methods of acquiring knowledge, presenting knowledge, classifying and compiling knowledge. УМ1 - Ability to apply methods and tools of modern information technologies for the development of automated design systems and their components in various fields. УМ4 - The ability to use knowledge of decision support methods and the ability to use them in the automated design of complex objects and systems AiB3 - The ability to take responsibility for the work performed, to make decisions independently, to achieve the set goal in compliance with the requirements of professional ethics
Required prior and related subjects: Previous educational disciplines: System programming, Methods and tools of computer information technologies, Algorithmization and programming. Associated and subsequent academic disciplines: Artificial intelligence systems.
Summary of the subject: The course "Methods and tools of machine learning" is developed for master degree students of Lviv Polytechnic National University. The basis of the course is to master the theoretical foundations of methods and tools used in machine learning, develop students' skills for a better understanding of the basic ideas underlying these methods and tools, to teach students how to work with software that implements algorithms of machine learning. The main programming language for implementing machine learning algorithms in this course is Python. At the end of the course, a student will know: the main methods and algorithms of machine learning, methods of minimizing the average risk of machine learning, the peculiarities of supervised/unsupervised machine learning, evaluate the quality of machine learning, to solve the applied problems associated with the application of machine learning algorithms.
Опис: Topic 1. Basics of machine learning Definition and basic concepts. Main tasks. Application and problems. Learning scenarios. Algorithm selection rules. Stages of development of computer training programs. Software for use in computer-based learning. Topic 2. Method of classification of k-nearest neighbors The principle of operation of the k-nearest neighbors algorithm. Weighted and unweighted voting. An example of the algorithm. Advantages and disadvantages of the algorithm. An example of classification. Checking the classifier. Topic 3. Decision trees Representation of the decision tree. Tasks for solving with decision trees. Advantages and disadvantages of the algorithm. Fields of application. Entropy. Information growth. An example of using the tree construction algorithm. Problems with the information growth criterion. Retraining the algorithm. Topic 4. Linear regression Regression analysis. Using regression analysis. Types of regression. Choosing a regression model. The method of least squares. An example of using the method of least squares. Topic 5. The principle of regularization and the regression problem Causality, regression, correlation. Characteristics of regression types. Correlational dependencies. Application of correlation-regression analysis. Multiple regression. Assessment of the materiality of the relationship. Interpolation. Regularization. Approximation. Function approximation. Topic 6. Bayesian theory of classification Bayesian approach. Bayesian machine learning. Probability theory. Combinations. Permutations. Bayes theorem. Bayes formula. Application of Bayes theorem to machine learning. Topic 7. Support vector machines Linear classifier. The method of Lagrange multipliers. Application of the method of Lagrange multipliers for the method of support vectors. Practical implementation of the support vector machines method. Topic 8. Core Method of support vectors. Soft and hard borders. Types of nuclei. Core selection. Dimensionality reduction problem. An example of using MOV for classification in Python. Topic 9. Boosting Hypothesis about boosting. Ensemble. Bagging. Boosting algorithm. Adaptive boosting. Gradient boosting. Algorithm for building gradient boosting. Example of implementation of gradient boosting. Topic 10. Random forest Definition of random forest. Algorithm for building a random forest. Advantages and disadvantages of the algorithm. Fields of use of random forest. Topic 11. The k-means clustering method The k-means algorithm. Building a clustering algorithm. The main steps of the clustering algorithm. Advantages and disadvantages of the algorithm. An example of using the algorithm. Topic 12. Neural networks in computer training Biological prototype. Artificial neuron. Types of artificial neural networks. The validity of the use of neural networks. Description of the components and operation of neural networks. Training of artificial neural networks. Learning algorithms. Perceptron. Topic 13. Convolutional neural networks Application of convolutional neural networks. Concept of convolution for 1D and 2D array of data. Algorithm for determining two-dimensional convolution. Implementations of the convolution algorithm. Topic 14. Pattern recognition using computer learning algorithms The principle of the scanning window. Viola-Jones algorithm. Training classifiers. Used in the boosting algorithm and AdaBoost development. Recognition of 3D objects. Algorithms for comparing 2D and 3D objects. Topic 15. Search systems and their principle of action Basic definitions. Types of search engines. Mechanisms of search algorithms. Advantages and disadvantages of search algorithms.
Assessment methods and criteria: Assessment of students' knowledge in the discipline "Methods and tools for machine learning" is carried out in accordance with the working curriculum in the form of a semester control, which is carried out at the end of the semester and includes the results of the current control of students' knowledge, which is assessed for the performance of laboratory work, and a control measure - the answer to the corresponding exam ticket. The control measure is a mandatory type of control and is conducted in written and oral form at the end of the semester. Current monitoring of lecture classes is carried out in order to identify the student's readiness for classes in the following forms: • a selective oral survey before the start of classes; • evaluation of the student's activity in the course of classes, submitted proposals, original solutions, clarifications and definitions, additions to previous answers, etc. Control questions are divided into: a) test tasks - choose the correct answers; b) problematic – creation of problematic situations; c) questions-replies - to identify cause-and-effect relationships; d) situational tasks - to determine the answer according to a certain situation; e) issues of a reproductive nature - determination of practical significance.
Критерії оцінювання результатів навчання: Methods of knowledge diagnosing: 1. Written exam. 2. Laboratory works. 3. Presentation and defense of the project. Criteria for evaluation: Total for discipline - 100 points, among them: Laboratory works (5) - 25 points Control work (1) - 20 points Exam (1) - 55 points
Порядок та критерії виставляння балів та оцінок: 100–88 points – (“excellent”) is awarded for a high level of knowledge (some inaccuracies are allowed) of the educational material of the component contained in the main and additional recommended literary sources, the ability to analyze the phenomena being studied in their interrelationship and development, clearly, succinctly, logically, consistently answer the questions, the ability to apply theoretical provisions when solving practical problems; 87–71 points – (“good”) is awarded for a generally correct understanding of the educational material of the component, including calculations, reasoned answers to the questions posed, which, however, contain certain (insignificant) shortcomings, for the ability to apply theoretical provisions when solving practical tasks; 70 – 50 points – (“satisfactory”) awarded for weak knowledge of the component’s educational material, inaccurate or poorly reasoned answers, with a violation of the sequence of presentation, for weak application of theoretical provisions when solving practical problems; 49-26 points - ("not certified" with the possibility of retaking the semester control) is awarded for ignorance of a significant part of the educational material of the component, significant errors in answering questions, inability to apply theoretical provisions when solving practical problems; 25-00 points - ("unsatisfactory" with mandatory re-study) is awarded for ignorance of a significant part of the educational material of the component, significant errors in answering questions, inability to navigate when solving practical problems, ignorance of the main fundamental provisions.
Recommended books: 1. Andrushchak N.A. Methods and tools of machine learning: lecture notes for students of the second (master's) level of higher education of the Institute of Computer Science and Information Technology. - Lviv: Lviv Polytechnic National University Publishing House, 2018. - 224 p. (in Ukrainian) 2. Andrushchak N.A. Methods and tools of machine learning: laboratory workshop for students of the second (master's) level of higher education at the Institute of Computer Science and Information Technology. - Lviv: Lviv Polytechnic National University Publishing House, 2018. - 125 p. (in Ukrainian) 3. Mitchell T. Machine learning / T. Mitchell. – McGraw Hill, 1997. 4. Hastie T. The elements of statistical learning / T. Hastie, R. Tibshirani, J. Friedman. – Springer, 2001. 5. Ripley B. D. Pattern recognition and neural networks / B. D. Ripley. – Cambridge University Press, 1996. 6. Bishop C. M. Pattern recognition and machine learning / C. M. Bishop. – Springer, 2006. 7. Duda R. O. Pattern classification / R. O. Duda, P. E. Hart, D. G. Stork. – New York : JohnWiley and Sons, 2001.
Уніфікований додаток: Lviv Polytechnic National University ensures the realization of the right of persons with disabilities to obtain higher education. Inclusive educational services are provided by the Service of accessibility to learning opportunities "Without restrictions", the purpose of which is to provide permanent individual support for the educational process of students with disabilities and chronic diseases. An important tool for the implementation of the inclusive educational policy at the University is the Program for improving the qualifications of scientific and pedagogical workers and educational and support staff in the field of social inclusion and inclusive education. Contact at: St. Karpinsky, 2/4, 1st floor, room 112 E-mail: nolimits@lpnu.ua Websites: https://lpnu.ua/nolimits https://lpnu.ua/integration
Академічна доброчесність: The policy regarding the academic integrity of the participants of the educational process is formed on the basis of compliance with the principles of academic integrity, taking into account the norms "Regulations on academic integrity at the Lviv Polytechnic National University" (approved by the academic council of the university on June 20, 2017, protocol No. 35).