Mashg‘ulotlar shakli: ma’ruza (M)
|
|
Introduction. Goals and objectives of the discipline "machine learning".
Basic terms of machine learning algorithms. Application of machine learning algorithms in real areas (image classification, spam detection, recommendation systems, natural language processing). The role of machine learning in artificial intelligence. Elements of the machine learning algorithm in modern artificial intelligence applications.
|
|
Types of machine learning.
Supervised (supranational) training. Unsupervised (unmanaged) learning. Semi-supervised (semi-free) learning. Algorithms of enhanced (repeated) learning. Algorithms of self-learning (self-control). Relocated (transfer) training. Online training. Package training.
|
|
The main processes that make up machine learning.
Data collection and processing. Function engineering. Model selection. Training and optimization. Checking and evaluating the model.
|
|
Tools and libraries used in machine learning.
Necessary tools. The Python software environment. Basic actions. Python custom machine learning libraries (scikit-learn, TensorFlow, PyTorch). Working with data reading functions for the model.
|
|
Tools and libraries used in machine learning.
Methods of graphical representation of data. Understanding the importance of graphical representation in machine learning, description and functions of Python custom graph-forming libraries (matplotlib, Pandas Plotting, Plotly).
|
|
Training sample (data set).
Methods of building a training sample in machine learning. Methods of creating, collecting and preprocessing a training sample. Generating a training sample
|
|
Training sample (data set).
Training sample generation functions (Pandas package). Working with existing training samples (open dataset). Open source learning options.
|
|
The problem of linear regression in machine learning.
The concept of linear regression. Building a linear regression model. Coefficients of the regression model. Calculation of the Cost function. Checking the accuracy of the model. Linear regression with one variable. A multidimensional view of regression analysis. Building a multidimensional linear regression model. The gradient descent method. Stochastic gradient descent
|
|
Classification issues in machine learning.
The concept of classification. Classification as one of the main approaches to machine learning. The question of learning. The concept of logistic regression. Building a logistic regression model. Calculation of probability values and determination of decision limits in the classification process (decision boundary). The Softmax regression function.
|
|
The concept of regularization.
Optimization issues. Regularization of the L1 and L2 levels in the learning process. Linear methods of data organization (Regularized linear models). Regulation of linear and logistic regression.
|
|
Classical classification algorithms.
Classification in machine learning description of naive Bayes and K-Nearest Neighbor algorithms. Random forest and decision tree algorithms. The principle of classification of the vector support machine algorithm. Advantages and disadvantages of machine learning classification algorithms
|
|
Algorithms for teaching without a teacher (unaccompanied).
The concept of teaching without a teacher. Tasks solved using unsupervised learning algorithms (clustering, scaling down, anomaly detection). Classical algorithms of teaching without a teacher. Features of the dataset for unsupervised learning algorithms.
|
|
Solving the clustering problem.
Basic concepts of basic clustering methods and algorithms, solving the problem of clustering a dataset using hierarchical clustering, K-means and other methods. Solving clustering problems in Python.
|
|
Evaluation and selection of models.
An estimated metric. Indicators for evaluating the effectiveness of the model. Evaluation indicators for the classification task (accuracy, Precision, recall, F1-score, Confusion matrix).
|
|
Evaluation and selection of models.
Estimated indicators for the regression task (MAE, MSE). Estimated indicators on the clustering problem. The concepts of cross-validation, bias, and variance in the evaluation of models.
|
|
Artificial neural networks.
The concept of a biological and artificial neuron. The concept of a neural network. Performing logical actions on neurons. The task of the perceptron concept. Schematic representation of the principle of operation of an artificial neural network. Building a simple neural network. Weight coefficients and their calculation.
|
|
Artificial neural networks.
The levels of the neural network. Activation functions, solving the binary classification problem using a simple neural network. Building multi-level neural networks. Incoming, outgoing and hidden levels in a multi-level neural network. Actions to update the values of weight coefficients in a multilevel neural network (backpropogation). Solving the problem of regression and classification through a multilevel perceptron.
|
|
Building a neural network based on software tools.
Functions for building a neural network in the Python environment and using existing libraries. Creating and printing a neural network model. The problem of neural network training. The question of learning in simple and complex neural networks.
|
|
Building a neural network based on software tools.
Choosing a training set for neural network training. Preprocessing of the training sample. Custom libraries for building Python neural networks (numpy, TensorFlow, PyTorch) and custom environments (Google Colab, etc.).
|
|
Solving regression and classification problems based on a neural network.
Calculation of the functions of the Loss function and Gradient Descent in a neural network. Improving the accuracy of the model for regression and classification.
|
|
The concept and types of advanced training.
The concept of deep learning. Solving artificial intelligence problems through in-depth training. The stages of advanced training. Types of advanced training.
|
|
The concept and types of advanced training.
The capabilities of CN, RN, LSTM, DFF and other deep learning algorithms.
|
|
