This seminar will provide you with the knowledge and ability to design, code and evaluate machine learning models.

Introduction to Machine Learning

Machine Learning (ML) is the field that deals with designing algorithms that learn from examples. ML gained an incredible popularity in recent years, due to its ability to review vast amounts of data. ML provides extraordinary value for a variety of tasks, ranging from spam filtering to machine translation.

The most common ML tasks are regression (prediction) and classification, both of which are supervised learning tasks. Countless system employ ML models including financial analysis, advertisement, document classification and topic analysis, object detection in images, detection and mitigation of cyber-attacks and so forth. This seminar focuses on these types of supervised ML.

Supervised ML requires a dataset with known labels (the desired outputs of the model). It is typically used for 2 main use cases:

  1. For improving the performance of a system based on large amounts of data.
  2. For automating tasks that require scale, consistency or velocity with predictive capabilities that are not available in static software.

Models are applied anywhere between micro-second decisions with very few indicators, to large-scale systems that process terabytes of data every day.

What you will learn

The necessary coding skills: You will learn basic Python and how to use the common toolboxes to be able to design, train and apply ML models.

  • How to validate an ML model: You will know the pitfalls of validating an ML model and how to avoid them. You’ll be able to predict the performance of your algorithm before you apply it.
  • Building models: You will learn to take a model from the level of setting goals to the point you have a working, validated model. You’ll implement a model end-to-end during the seminar.

Target Group

  • Professionals in technical fields who would like to add ML to their toolbox.
  • Project managers and group leader who have to bear responsibility for the content and have to assess rules procedures in detail.

About the teacher

Dr. Liron Allerhand received his Ph.D. in engineering from Tel-Aviv university in 2014, held senior ML and algorithmic positions in industry both in Germany and in Israel and is currently a senior applied researcher in Microsoft where uses machine learning to protect companies from cyber-attacks.

Main Topics

This seminar consists of three modules, which will be dealt with in three days:

Part 1: Python and Preliminaries

Contents are:

  • Basic Python
  • Python toolboxes: Numpy and Pandas
  • Types of learning algorithms and their applications

During the exercise sessions you will write simple Python scripts.

Part 2: ML theory and evaluating ML models

Contents are:

  • Defining a proper objective for an ML model
  • Strong vs. weak learning algorithms
  • Overfitting and underfitting, the bias-variance trade-off
  • The phases of an ML project
  • How to lie with statistics, and how to keep statistics from lying to us
  • Loss functions: Cross entropy, Hinge loss, L2 loss, L1 loss
  • Logistic regression, linear regression and how they work

In the exercises, you design a methodology to correctly approach an ML problem.

Part 3:  Building ML models

Contents are:

  • Exploring data
  • Features vs. raw data
  • Cleaning data, handling missing data and outliers
  • Transforming and scaling features
  • Regularization

In the exercises, you will design, implement and validate an ML model end-to-end.

This seminar will be given in English

Termin: tbd, 3 days

Location: In-house or Campus University of Stuttgart (Stuttgart-Vaihingen)

Necessary equipment: Laptop

Preliminary knowledge: Basic coding skills in some high level scripting language like Matlab, Python, or R are desired.

Cost: 1950 € (zzgl. MwSt.)

Lecturer: Dr. Liron Allerhand (Microsoft Israel R&D Center)

Please observe our terms and conditions regarding the seminars (german only).

Open questions? Please contact us.
Dieter Schwarzmann
Tel: +49 1515 6026850

Seminar vormerken

    Die von Ihnen zur Verfügung gestellten Angaben werden zur Bearbeitung Ihres Anliegens und Beantwortung Ihrer Anfrage gespeichert. Daten werden nicht an Dritte weitergegeben. Datenschutzerklärung
    Durch das Anklicken der Box akzeptieren Sie diese Nutzungsbedingungen.

    Open questions? Please contact us.
    Dieter Schwarzmann
    Tel: +49 1515 6026850