fbpx




Advanced Machine Learning Training

Advanced Machine Learning Training
 

Due to the COVID-19 our training courses will be taught via an online classroom.

Receive in-depth knowledge from industry professionals, test your skills with hands-on assignments & demos, and get access to valuable resources and tools.

This course offers a deep-dive into advanced machine learning concepts such as Bayesian learning, ensemble models, decision & regression trees, and optimization. The lessons that are presented here focus not only on the mathematical background of those concepts but also on real-life applications. After this course, you will be able to design advanced machine learning and optimization algorithms in python. This course is ideal for data scientists who want to take the next step into machine learning. As requirements, strong experience with python and general machine learning knowledge are needed. 

Are you interested? Contact us and we will get in touch with you.

 

Get in touch for more information

Fill in the form and we will contact you about the Advanced ML training:

Academy: Advanced AI
I agree to be contacted *

About the training & classes

The Advanced Machine Learning training is split in 4 days. Click below to see a detailed description of each class: 

Decision & Regression Trees

In this training you will be introduced to Decision Trees, how they are constructed, and what algorithms you need.

You will learn how to optimize them and why this is necessary. We will also explain how these concepts are extended to their regression equivalents, using regression trees. In the end, we will implement a decision tree algorithm from scratch and then apply a scikit-learn decision tree to the Shuttle dataset provided by NASA.

The training includes theory, demos, and hands-on exercises.

After this training you will have gained knowledge about:

  • Understanding high dimensional data
  • Decision Tree basics
  • Decision nodes, leaf nodes
  • Entropy, Information Gain and Gini impurity
  • ID3, CART algorithms
  • Optimizing decision trees
  • Regression trees
  • Advantages and disadvantages
  • Lab sessions to get hands-on experience applying this knowledge
Problem Solving & Optimization

Problem Solving & Optimization encompasses a broad range of techniques and subfields, that can be applied to many real-world applications.

This training aims to provide a thorough overview of these techniques and valuable knowledge about how they can be applied, in machine learning as well as beyond. We will provide an overview of different groups of optimization methods, from ‘exact’ methods such as Mathematical Programming and Gradient-Based Optimization to heuristics methods like Simulated Annealing and Evolutionary Computation, spending more time on the most important ones. Finally, we discuss the most common challenges in optimization, e.g. local vs global optima, the exploration vs exploitation tradeoff and tuning, before concluding with some practical notes on specialized solvers and good use cases.

The training includes theory, demos, and hands-on exercises.

After this training you will have gained knowledge about:

  • Problem solving contexts
  • Solution representations, constraints and objective functions
  • Mathematical programming
  • Gradient-based optimization
  • Black box optimization
  • (Meta)heuristics
  • Simulated Annealing
  • Tabu search
  • Evolutionary Computation
  • Local vs global optima
  • Exploration vs Exploitation tradeoff
  • Tuning
  • Specialized solvers
  • No free lunch theorem
Bayesian Learning

The training aims to give an introduction into Bayesian statistics.

We explain how Bayesian Statistics compare to a frequentist approach, and dive into the calculations needed to create Baysian models. We will also have a first look at the domain of probabilistic programming. In the theoretical part, we introduce the Bayesian formula and give an introduction into different types of statistical distributions (Bernoulli, Poisson, Uniform, Normal, Exponential).

With this theoretical basis, we will walk through the calculation of Bayesian updates, Bayesian parameter estimation and the working of a Naïve Bayes classifier.

After this theoretical part, we dive into the application of the theory in the lab. The lab has three sections: 1. Coin Tossing. We take a biased coin, and explore the idea of Bayesian updates to describe the evolution of our prior beliefs about the coin. 2. Diabetes dataset, single variable. We will implement a naïve Bayesian classifier for a single variable. 3. Multivariable Predictions. We will extend the single variable prediction to a multivariable model.

There is also an additional lab that illustrates how probabilistic programming works, and that introduces the use of the pymc3 library.

After the training, students will have gained knowledge about:

  • The definitions of prior, likelihood, posterior and evidence
  • The definition and application of Bayes Rule
  • How to iteratively update prior distributions from new observations with Bayesian Inference
  • How a Gaussian Naïve Bayes algorithm can be implemented, both for single and multiple variables.
  • How probabilistic modelling looks, and how pymc3 can be used to make probabilistic models for parameter estimation.
Ensemble Models

As Machine Learning encompasses a large set of different algorithms, many of them (if not all) suffer from high bias or variance.

Ensemble Learning aims to reduce bias and/or variance using methods such as bagging, boosting and stacking, thereby combining weak learners into stronger ones. In this training, we first revisit the Bias-Variance Tradeoff and give a good motivation for how Ensemble Learning tries to address this. We then discuss Bootstrap Aggregating (bagging), its role in reducing variance and how it is implemented in Random Forests. Continuing to Boosting, we explain how it aims to tackle issues of too high bias and discuss implementations like Adaboost and Gradient Boosting. We address how model performance can be improved using Stacking, and when this generally works best. We conclude with an overview of techniques and their advantages/disadvantages.

Having learned the theory, we apply these methods in practice during a lab exercise, thereby giving more understanding about all three methods, i.e. Stacking, Bagging and Boosting.

The training includes theory, demos and hands-on exercises.

After this training you have gained knowledge about:

  • Combining algorithms
  • Bias-Variance Trade-off
  • Bagging (bootstrap aggregating)
  • Majority Voting
  • Random Forests
  • Boosting
  • Adaboost & Gradient Boosting
  • Stacking