Training agenda
- Introduction to ML
- Definition of ML
- ML vs. artificial intelligence
- Applications of ML
- Skills needed in ML
- Structured/unstructured data
- Features/instances/labels/predicted variable
- Classification/regression/clustering
- Types of ML
- Training/prediction/evaluation
- Splitting data into training and test sets
- Linear regression
- Logistic regression
- Algorithm vs. model
- ML workflow
- ML Frameworks
- Hello world of ML – iris classification
- Data preparation
- Problems with data
- Basics of data exploration and visualization
- Feature types
- Feature encoding
- Standardization and scaling
- Outliers
- 3-sigma rule
- Dealing with missing data
- Feature selection/dimensionality reduction
- Selected problems in ML
- Model evaluation
- Overfitting and underfitting
- Randomness and reproducibility
- Cross-validation
- Parameters vs. hyperparameters
- Hyperparameter optimization
- Classical ML algorithms
- Taxonomy of ML algorithms
- K-nearest neighbors algorithm
- Decision tree
- Ensembling
- Random forest
- Multiclass and multilabel classification
- Clustering: k-means algorithm
- Artificial neural networks
- Motivation and biological inspiration
- Neuron model
- Activation functions
- Multilayer perceptron – architecture
- Multilayer perceptron – prediction
- Multilayer perceptron – learning (backpropagation)
- Image encoding
- Batch learning
- Types of neural networks
- Interpretability of model
- Further learning
- Course summary: questions/comments/discussion