AI specialist and mathematical modeling

Introduction to Machine Learning (approx. 5 days)

Why machine learning?

Application examples

Supervised learning, unsupervised learning, partially supervised learning, reinforcement learning

Examples of data sets

Getting to know data

Training, validation and test data

Viewing data

Making predictions

Supervised learning (approx. 5 days)

Classification and regression

Generalization, overfitting and underfitting

Size of the data set

Algorithms for supervised learning

Linear models

Bayes classifiers

Decision trees

Random Forest

Gradient Boosting

k-nearest neighbors

Support Vector Machines

Conditional Random Field

Neural Networks and Deep Learning

Probabilities

Unsupervised learning (approx. 5 days)

Types of unsupervised learning

Preprocessing and scaling

Data transformations

Scaling training and test data

Dimension reduction

Feature engineering

Manifold learning

Principal component decomposition (PCA)

Non-negative matrix factorization (NMF)

Manifold learning with t-SNE

Cluster analysis

k-Means clustering

Agglomerative clustering

Hierarchical cluster analysis

DBSCAN

Cluster algorithms

Evaluation and improvement (approx. 2 days)

Model selection and model evaluation

Tuning the hyperparameters of an estimator

Cross-validation

Grid search

Evaluation metrics

Classification

Project work (approx. 3 days)

To consolidate the content learned

Presentation of the project results

Introduction to Deep Learning (approx. 1 day)

Deep learning as a type of machine learning

Fundamentals of neural networks (approx. 4 days)

Perceptron

Calculation of neural networks

Optimization of model parameters, backpropagation

Deep learning libraries

Regression vs. classification

Learning curves, overfitting and regularization

Hyperparameter optimization

Stochastic gradient descent (SGD)

Momentum, Adam Optimizer

Learning rate

Convolutional Neural Network (CNN) (approx. 2 days)

Image classification

Convolutional layers, pooling layers

Reshaping layers, flattening, global-average pooling

CNN architectures ImageNet-Competition

Deep neural networks, vanishing gradients, skip connections, batch normalization

Transfer Learning (approx. 1 day)

Adaptation of models

Unsupervised pre-training

Image data augmentation, explainable AI

Regional CNN (approx. 1 day)

Object localization

Regression problems

Branched neural networks

Methods of creative image generation (approx. 1 day)

Generative Adversarial Networks (GAN)

Deepfakes

Diffusion models

Recurrent neural networks (approx. 2 days)

Sequence analysis

Recurrent layers

Backpropagation through time (BPTT)

Analysis of time series

Exploding and vanishing gradient problems

LSTM (Long Short-Term Memory)

GRU (Gated Recurrent Unit)

Deep RNN

Deep LSTM

Text processing using neural networks (approx. 2 days)

Text preprocessing

Embedding layers

Text classification

Sentiment analysis

Transfer learning in NLP

Translations

Sequence-to-sequence method, encoder-decoder architecture

Language models (approx. 1 day)

BERT, GPT

Attention layers, Transformers

Text generation pipelines

Summarization

chatbots

Deep reinforcement learning (approx. 1 day)

Control of dynamic systems

Agent systems

Training through rewards

Policy Gradients

Deep Q-learning

Bayesian neural networks (approx. 1 day)

Uncertainties in neural networks

Statistical evaluation of forecasts

Confidence, standard deviation

Unbalanced data

Sampling methods

Project work (approx. 3 days)

To consolidate the content learned

Presentation of the project results

MATLAB basics (approx. 2 days)

MATLAB user interface

Reading data from a file

Variables, arrays, operators, basic functions

Graphical representation of data

Customizing diagrams

Exporting graphics

Variables and commands (approx. 2 days)

Relational and logical operators

Sets, sets with 2D solids (polyshape)

Performing mathematical and statistical calculations with vectors

Graphics in statistics

Analysis and visualization (approx. 1 day)

Creating and modifying matrices

Mathematical operations with matrices

Graphical representation of matrix data

Matrix applications: Mappings, rotation, systems of linear equations, least square method

Artificial intelligence (AI) in the work process

Presentation of specific AI technologies

and possible applications in the professional environment

Data processing (approx. 1 day)

Data types: Structure arrays, cell arrays, string vs. char, categorical, datetime and much more.

Creating and organizing tabular data

Conditional data selection

Importing/exporting with Matlab: folder structures, .mat data, table data, continuous texts

MATLAB programming (approx. 3 days)

Control structures: loops, if-else, exceptions

Functions

Object-oriented programming

App design

Simulation in MATLAB (approx. 5 days)

Numerical integration and differentiation

Basics of simulation of ordinary differential equations, Matlab ODE and solver options

Simulation technology in Matlab: input parameters, data interpolation, simulation studies

Simulation control: event functions (zero crossing), output functions

Application examples, e.g. simulation of an electric motor, simulation of a rocket

Simulink (approx. 4 days)

Basics of Simulink: Diagrams, functions, signals and differential equations

Functions, subsystems and libraries

Import/export, lookup tables, control

Zero-crossing, automation of simulation tasks (Matlab access)

Application examples, e.g. simulation of an aircraft drive train

Project work (approx. 2 days)

To consolidate the content learned

Presentation of the project results