Webinars: BASICS Machine Learning with Python
YOUR QUICK START!
Dates: April 15–18, 2024; June 10–13, 2024; and October 7–10, 2024
Webinar 1: Python Basics; Data Preparation with NumPy and pandas
Python has now become the standard in the field of machine learning. But not Python alone: rather, it is the packages NumPy, pandas, scikit-learn, and matplotlib that have made this possible.
NumPy provides arrays for handling data structures typical in machine learning. Many packages used in machine learning require NumPy objects. If more heterogeneous data structures are to be used (i.e., comparable to relational tables in databases), pandas is the de facto standard. The seminar begins with an overview of Python. Programming knowledge of a programming language is assumed.
Contents :
-
Python Basics, Built-in Types, Data Structures, Sequences, Control Flow, Functions
- NumPy Basics: Arrays and Vectorized Computation (Indexing, Slicing)
-
pandas Data Structures: Series, DataFrame, Index Objects including Loading, Storage and File Formats
Webinar 2: Data preparation with pandas, visualizations with matplotlib.
The seminar builds on the first webinar (Python Basics; Data Preparation with NumPy and pandas). With pandas, heterogeneous data structures (comparable to relational tables in databases) can also be transformed. The matplotlib package allows for quick visualization of the data. The seaborn package offers a unique feature: it uses both pandas and matplotlib to quickly create graphics.
Contents :
Data Clearing and Preparation with pandas
Merging multiple data sets: Join and Combine
Data Aggregations and Group Operations with pandas: Reshape, Aggregations , Grouping
matplotlib API Primers, Figures and Subplots
Colors, Markers, Line Styles, Ticks, Labels, Legends, Annotations and Drawings, Saving Plots to a File
Plotting with pndas and seaborn: Line Plots, Bar Plots, Histograms and Density Plots, Scatter or Point Plots
Webinar 3: Basics of Supervised Learning with scikit-learn
In supervised learning, a model is trained to predict a target variable (e.g., purchase/non-purchase, customer churn (yes/no), purchase value) based on suitable input variables (features). Depending on the target variable, the model is referred to as regression (purchase value) or classification (purchase/non-purchase). These models can be conveniently trained using the scikit-learn package. Model training should be monitored for overfitting: an algorithm can use the training data to perfectly predict the target variable, but when new data is introduced, it often performs less well. This trade-off must be controlled to ensure models can also be applied to new data.
Contents :
Basics of scikit-learn
Using pandas and numpy data for model training
Typical steps in scikit-learn (regression): importing the necessary submodules, creating an object, fitting the data, checking the results (overfitting) with appropriate fit measures, making predictions with the trained model;
Classification: differences from regression, fit measures, outcome control;
Saving and deploying the trained models;
Webinar 4: Deep Learning with Keras and TensorFlow including Notes on Reinforcement Learning
Deep learning refers to deep neural networks, or more simply, complex models that attempt to mimic the information processing of living organisms. Deep learning models can also be used to train regression and classification models (as described in Module 3). Deep learning models have the advantage that, on the one hand, the data volumes can be larger and more complex: processing images in the form of 3D arrays (e.g., NumPy ndarrays) is easily possible. Keras and TensorFlow, among others, have established themselves in this area.
Contents :
Basics of deep learning models including network topology and different model types;
Regression with Keras and TensorFlow
Image classification with Keras and TensorFlow
Reinforcement Learning