Machine Learning Practitioner E-Learning Training

Met de ML Practitioner Learning Kit ontwikkel je praktische, technische vaardigheden om machine learning modellen te ontwerpen, optimaliseren en implementeren in realistische productieomgevingen. Deze Learning Kit is bedoeld voor data scientists, machine learning engineers en analytics practitioners die verder willen gaan dan basiskennis en productiegerichte ML-skills willen opbouwen.

Demo Machine Learning Practitioner Training

De focus ligt op de praktische onderdelen die bepalen of een machine learning model succesvol wordt in productie. Je werkt met data preparation, feature engineering, anomaly detection, hyperparameter tuning, MLOps, model deployment en reinforcement learning.

Door de vijf onderdelen leer je een machine learning model end-to-end beheren: van ruwe data naar schone, model-ready datasets, vervolgens naar training, tuning, deployment, monitoring en geavanceerde technieken zoals reinforcement learning.

Wat leer je in deze Learning Kit?

• Data voorbereiden voor machine learning modellen
• Missing values en outliers herkennen en behandelen
• Categorical data encoden en numeric features schalen
• Feature engineering pipelines bouwen met scikit-learn
• Feature selection en dimensionality reduction toepassen
• Hyperparameters optimaliseren met moderne tuning tools
• Anomalies, fraud, errors en rare events detecteren
• MLOps-principes toepassen voor model deployment en lifecycle management
• Modellen trainen, registreren en deployen met MLflow op Databricks
• Reinforcement learning concepten begrijpen en toepassen

Voor wie is deze Learning Kit bedoeld?

Deze Learning Kit is geschikt voor:

• Data Scientists
• Machine Learning Engineers
• Analytics Practitioners
• Data Analysts die willen doorgroeien naar machine learning
• AI Developers
• MLOps Engineers
• Data professionals die modellen production-ready willen maken
• Professionals die praktische ML-vaardigheden willen ontwikkelen

Deze LearningKit met meer dan 21 leeruren is verdeeld in drie sporen:

Cursusinhoud

Onderdeel 1: Feature Engineering for ML Models

Dit onderdeel legt de basis voor goed presterende machine learning modellen door te focussen op data preparation en feature engineering. Je leert missing values en outliers behandelen, categorical variables encoden en numeric features schalen zodat datasets geschikt zijn voor modeltraining.

Handling Missing Values and Outliers in Data

Course: 1 Hour, 22 Minutes

• Course Overview
• Data Preparation
• Missing Values and Outliers
• Outlier Identification and Analysis
• Outlier Detection and Management
• Loading and Preparing Data for Cleaning
• Dropping Records and Imputing Values to Address Missing Data
• Implementing Multivariate Imputation
• Detecting and Visualizing Outliers with Box Plots
• Identifying and Computing Outliers Using IQR
• Capping Outliers with IQR and Evaluating Results
• Using the Z-Score Technique for Outlier Detection and Capping
• Course Summary

Encoding Categorical Data for Machine Learning

Course: 54 Minutes

• Course Overview
• Data Types and Processing Techniques
• One-Hot Encoding of Categorical Data
• Encoding Methods for Categorical Data
• Techniques for Encoding Categorical Variables
• Performing Label Encoding of Categorical Values
• Performing One-Hot Encoding of Nominal Categorical Variables
• Encoding Categorical Variables with the Ordinal Encoder
• Discretizing Variables and Training a Random Forest Model
• Course Summary

Scaling Numeric Data for Machine Learning

Course: 37 Minutes

• Course Overview
• Scaling Numeric Features
• Scaling Techniques and Their Applications
• Applying Min-Max Scaling to Numeric Features
• Implementing Standard Scaling and Frequency Encoding
• Training a KNN Classifier and Evaluating Performance
• Course Summary

Feature Engineering Techniques for Machine Learning

Course: 1 Hour, 7 Minutes

• Course Overview
• Feature Engineering Features and Mitigating Overfitting Risks
• Feature Creation and Model Impact
• Synthesizing Features and Constructing Scikit-Learn Pipelines
• Executing Pipeline Fit, Transformations, and Model Evaluation
• Performing Feature Engineering
• Modeling Non-Linear Relationships with Polynomial Features
• Creating Polynomial Features for a Regression Model
• Performing Log Transformations on Features
• Implementing Principal Component Analysis
• Applying Power Transformations and Performing PCA
• Course Summary

Feature Selection and Dimensionality Reduction

Course: 1 Hour, 11 Minutes

• Course Overview
• Feature Selection Methods
• Feature Selection Techniques
• Dimensionality Reduction with Principal Component Analysis (PCA)
• Analyzing Multicollinearity and Performing Feature Selection
• Performing Feature Selection Using Variance Threshold and the F-Statistic
• Applying Mutual Information Regression to Enhance Feature Selection
• Implementing Model-Based Feature Selection with Ridge Regression
• Executing Sequential Feature Selection
• Performing Classification Feature Selection Using Chi-Squared and F-Statistic
• Executing Recursive Feature Selection
• Course Summary

Onderdeel 2: Hyperparameter Tuning for Machine Learning

Dit onderdeel richt zich op hyperparameter tuning: een belangrijke factor voor betere modelprestaties. Je leert het verschil tussen model parameters en hyperparameters en onderzoekt technieken zoals grid search, random search en cross-validation.

Hyperparameter Tuning Techniques

Course: 1 Hour, 12 Minutes

• Course Overview
• Model Parameters and Hyperparameters
• Hyperparameter Tuning
• Hyperparameter Tuning and Early Stopping Techniques
• Cross-Validation to Mitigate Overfitting
• K-Fold Cross-Validation
• Decision Trees and Hyperparameter Tuning
• Balancing Hyperparameters in Decision Trees
• Regularization and Splitting in Decision Trees
• Hyperparameter Tuning Algorithms
• Hyperparameter Tuning Methods
• Course Summary

Hyperparameter Tuning with scikit-learn

Course: 1 Hour, 51 Minutes

• Course Overview
• Hyperparameter Tuning with scikit-learn
• Performing Exploratory Data Analysis and Visualizing Correlations
• Building a Baseline Regression Model and Evaluating Performance
• Tuning Decision Tree Hyperparameters
• Executing GridSearchCV to Tune Regression Models
• Utilizing GridSearchCV Results and Analyzing Best Models
• Implementing and Evaluating RandomizedSearchCV
• Employing HalvingGridSearchCV for Hyperparameter Tuning
• Using HalvingGridSearchCV with Dynamic Resources
• Selecting Optimal Regression Models Through Hyperparameter Tuning
• Building a Baseline Classification Model
• Implementing Hyperparameter Tuning with GridSearchCV for Classification Models
• Performing GridSearchCV with Multiple Scoring Metrics
• Exploring Randomized and Halving Random Search Strategies
• Course Summary

Hyperparameter Tuning with Hyperopt on Databricks

Course: 1 Hour, 57 Minutes

• Course Overview
• The Databricks Platform
• Hyperopt on Databricks
• Search Algorithms in Hyperopt
• Creating a Databricks Workspace on Azure
• Configuring and Launching an Apache Spark Cluster
• Creating a Databricks Volume and Uploading CSV Data
• Writing Python Code Using Databricks Notebooks
• Encoding and Scaling Data and Training a Model
• Tracking a Model with MLflow and Exploring Experiment Data
• Executing and Analyzing Hyperparameter Tuning with Hyperopt
• Analyzing Hyperparameter Tuning Runs in Databricks
• Performing Distributed Hyperparameter Tuning with SparkTrials
• Course Summary

Hyperparameter Tuning with Ray Tune on Databricks

Course: 50 Minutes

• Course Overview
• Key Features of Ray Tune
• Performing Hyperparameter Tuning with Ray Tune
• Understanding Ray Tasks and MLflow Integration
• Executing Parallel Model Training and Logging with Ray
• Executing Hyperparameter Tuning with Ray Tune
• Course Summary

Hyperparameter Tuning with Optuna

Course: 59 Minutes

• Course Overview
• Hyperparameter Optimization with Optuna
• Training a Baseline Classification Model and Exploring the Confusion Matrix
• Running an Optuna Study to Find the Best Hyperparameters
• Interpreting an Optuna Study
• Visualizing Hyperparameter Tuning Using Optuna Dashboard
• Hyperopt vs. Optuna vs. Ray Tune
• Course Summary

Automated Machine Learning with H2O AutoML

Course: 54 Minutes

• Course Overview
• H2O AutoML
• Training Regression Models with H2O AutoML
• Automatically Tuning Multiple Models Using H2O AutoML
• Training Models with Diverse Algorithms Using H2O AutoML
• Interpreting Models with H2O’s Explain Function
• Course Summary

Hyperparameter Tuning with Keras Tuner

Course: 47 Minutes

• Course Overview
• Hyperparameter Optimization with Keras Tuner
• Utilizing Hyperparameter Tuning with Keras Tuner in Colab
• Training a Convolutional Network for Image Classification
• Tuning Hyperparameters with Keras Tuner
• Visualizing and Evaluating Hyperparameter Tuning with TensorBoard
• Course Summary

Onderdeel 3: Anomaly Detection

Dit onderdeel leert je praktische technieken om outliers, fouten, fraudepatronen en zeldzame events te herkennen die de datakwaliteit en modelprestaties kunnen beïnvloeden.

Understanding Anomalies and Their Detection

Course: 49 Minutes

• Course Overview
• What Are Anomalies?
• Point Anomalies
• Contextual Anomalies
• Collective Anomalies
• Course Summary

Using Z-Scores and IQR for Anomaly Detection

Course: 47 Minutes

• Course Overview
• Anomaly Detection with Z-Scores
• Anomaly Detection with Interquartile Ranges (IQR)
• Anomaly Detection with Modified Z-Scores
• Using Interquartile Ranges to Identify Anomalies
• Using Z-Scores and Modified Z-Scores for Anomaly Detection
• Course Summary

Using LOF, iForest, and One-Class SVMs for Anomaly Detection

Course: 1 Hour, 43 Minutes

• Course Overview
• Anomaly Detection with Local Outlier Factor (LOF)
• Performing Anomaly Detection Using LOF
• Configuring the Sensitivity of LOF Using Number of Neighbors
• Anomaly Detection with Isolation Forest (iForest)
• Performing Anomaly Detection Using iForest
• Visualizing and Interpreting Anomalies Using Scatter Plots and Histograms
• Anomaly Detection with One-Class Support Vector Machines (OC SVMs)
• Performing Anomaly Detection with One-Class SVM and a Linear Kernel
• Performing Anomaly Detection with One-Class SVM and an RBF Kernel
• Anomaly Detection with Elliptic Envelope Detection
• Performing Anomaly Detection with Elliptic Envelope
• Course Summary

Onderdeel 4: MLOps and Model Deployment

Dit onderdeel overbrugt de kloof tussen modelontwikkeling en productie. Je leert hoe MLOps DevOps-principes uitbreidt naar machine learning, met aandacht voor data drift, model retraining, reproducibility, lifecycle management, CI/CD, infrastructure as code, containerization en automated testing.

MLOps and Model Deployment: Model Deployment and Serving Strategies

Course: 1 Hour, 18 Minutes

• Course Overview
• Model Deployment and Serving in MLOps
• Batch, Online, Streaming, and Edge Serving Modes
• Big Bang, Blue/Green, and A/B Deployment Strategies
• Model Hosting and Compute Choices
• Scalability Pattern and Update Frequency
• Model Servers and Managed ML Serving Platforms
• MLflow
• Information Captured in MLflow Runs
• Model Registry
• Course Summary

MLOps and Model Deployment: Contextualizing MLOps and DevOps

Course: 52 Minutes

• Course Overview
• Introducing MLOps
• CI/CD in MLOps
• IaC in MLOps
• Containerization and Orchestration in MLOps
• Automated Testing and Version Control in MLOps
• DevOps vs. MLOps
• Course Summary

MLOps and Model Deployment: Training and Deploying Models Using MLflow on Databricks

Course: 46 Minutes

• Course Overview
• Setting Up the Machine Learning Environment on Databricks
• Splitting and Preprocessing the Data for Machine Learning
• Training and Tracking Parameters and Metrics Using MLflow Runs
• Registering a Model with the MLflow Registry
• Deploying and Serving the Model
• Course Summary

Onderdeel 5: Reinforcement Learning

Dit onderdeel introduceert reinforcement learning: een machine learning paradigma waarbij agents leren door trial, feedback en reward in plaats van gelabelde data.

Implementing Simple Reinforcement Learning Methods

Course: 44 Minutes

• Course Overview
• Set Up an Environment for a Reinforcement Learning Agent
• Instantiating and Training an Agent
• Visualizing Metrics for a Trained Agent
• Setting Up an Environment and Agent for Shortest Path Computation
• Visualizing Metrics for the Trained Shortest Path Agent
• Course Summary

Contextualizing Reinforcement Learning

Course: 58 Minutes

• Course Overview
• Contextualizing Reinforcement Learning
• Actions and States
• Environment Modeling and MDP
• Policy Search
• Dynamic Programming and Q-Learning Intuition
• Q-Learning and SARSA
• Deep Q Networks
• Course Summary