Practical course Deep Learning for Python programmers
You are a Python programmer and want to learn the most important methods of the new topic Deep Learning and find out where you can integrate them into your current projects? Then this is the right course for you. This course covers step-by-step and in a hands-on way the basics of Deep Learning as well as the topics PyTorch Ecosystem, Pre-trained Models and Neural Networks. You will program these techniques, algorithms and methods with Python and thus get to know Deep Learning from a practical point of view. Learn more about this exciting Deep Learning practical course...
- Fast and hands-on introduction to Deep Learning with Python
- Decide if Deep Learning is suited for your projects
- Includes an introduction to the Python library for Deep Learning
- Learn basics to advanced Deep Learning topics in a practical way
Learn Deep Learning from our experts and you will be amazed at what is possible!
The course covers a selection of the following topics:
- Introduction to Machine Learning
- Introduction to Deep Learning
- Overview of the PyTorch Ecosystem
- Pre-trained models: Transformer, VGG, MobileNet
- Neural networks for different data types
- Developing models
For more details, please see the agenda below.
This Deep Learning course is suitable for people with very good Python knowledge, e.g. to the extent of our Python for Programmers course.
If you want to take this course with the R programming language, then check out the Deep Learning course with R. If you are a team leader, manager, executive, or non-programmer, check out our Artificial Intelligence for Decision Makers and Non-programmers basic course.
Individual: we specifically address your needs and take into account your previous knowledge, desired topics and focal points
Structured and easy to understand
Take your career, studies or training to the next level: with certificate
Safely and independently develop programmes (whether private, professional or for your start-up)
Lots and lots of practice: immediately applicable results
Small groups: max. 8 participants in the 3-day course max. 12 participants in online coaching
Developed by experts according to the Raed Method® & geared to the requirements of tech companies in 2020
E-mail support even after the end of the course
This course provides a practical introduction to modern deep learning techniques for students with Python skills who have no prior experience with machine learning. It covers all the tools, techniques, and processes necessary to successfully run simple machine learning projects with neural networks and provides a solid foundation for moving into more advanced topics.
This course can be run using PyTorch or TensorFlow as the deep learning library. The following description refers to the PyTorch version.
The table of contents lists the main topics covered. During the course, there are several hands-on workshops on different topics that are not listed separately in the table of contents. The course does not strictly follow the order in the agenda, as some topics are covered multiple times, typically with increasing difficulty.
What will you learn on the first day?
We start with an ML introduction. In this short section we cover questions like: What is ML? Which problems can be solved with ML techniques, which are less suitable? What does an ML development project look like? What best practices should you apply?
In many cases, the success of practical machine learning projects depends more on the quality of the data used to drive the algorithms than on the details of the ML algorithm. In this section, we'll cover the main tools for collecting, cleaning, analyzing, and preparing data in Python.
What will you learn on the second day?
On the second day, you'll learn Traditional Machine Learning. While Deep Learning is a major focus in academic research and provides superior results for most perceptual problems, traditional ML techniques are still viable solutions to many problems and are often used as baselines against which more elaborate solutions are evaluated.
What will you learn on day three?
In this part of the course, we will look at Deep Learning and in particular how pre-trained DL models can provide us with state-of-the-art solutions, even in cases where we have limited data or computational budget.
There are many opportunities for error when ML models are transitioned from training to production. While the topic is too broad to cover in detail in this course, we will at least provide an overview of some of the key concerns when transitioning smaller projects to production.
The Deep Learning course for Python programmers is taught by one of the following trainers:
Dr. Stefan Behnel
Expert: Python, Pytest, Unit Test und TDD, Clean Code, Clean Software Architektur, Fast Python, Cython
- Doctorate at the TU Darmstadt as Dr. Ing. in Software Architecture
References: 15 years of experience as a consultant, software developer and software architect in the financial services, automotive, publishing and tourism industries in the field of high-performance Python and open source, main developer of Cython, the data science library PANDA is based on Cython. Python training for Draeger, Apple, Sky Deutschland, IT companies, ...
Expert: Java, Python, Clean Code, Clean SW-Architecture, Refactoring, Testing, Train The Trainer
- Doctorate at LMU in the field of programming languages (2022).
References: 17 years of teaching and industry experience, thereof 12 years lecturer at the Ludwig-Maximilians-University Munich for Java, Python, Efficient Algorithms, Multiple awardsfor outstanding teaching at the LMU, book author for Java & soon Python at Springer and Orelly Verlag, developer of the RAED-Teaching Method®, Train the Trainer instructor, team training in Java and Python for BMW, VW, BA, SIEMENS, AGFA-Healthcare, TÜV Süd, Schufa AG, ..
Prof. Dr. Peer Kröger
Expert: Artificial Intelligence, Data Science, Big Data, SQL/NoSQL Database, Python, Java
- Doctorate at LMU in the field of Database and Data Science
References: Many years of practical experience in the implementation of data science projects as well as in consulting and training in the automotive industry, financial service providers and SMEs, among others. Approx. 150 peer-reviewed publications (cited over 8000 times) on the topic of data science, data mining, machine learning and AI. Member of the AI competence centre Munich Center for Machine Learning (MCML) at LMU Munich and professor for information systems and data mining at CAU Kiel.
Dr. Matthias Hölzl
- Doctorate at LMU in the field of Software Engineering
References: 30 years of teaching and industrial experience. Of which 18 years at Ludwig-Maximilians-Universität Munich, most recently as Professor for Software and Computational Systems Engineering. Training, technical coaching for machine learning, deep learning, process automation as well as review and improvement of software architecture in large IT projects. Python and Java trainings for Deutsche Bank, BMW, BA, VKB, etc. Editor and author of several books at Springer-Verlag and author of numerous scientific publications.
Of course, as a participant in the Deep Learning course for Python programmers you will receive a certificate. The prerequisite for this is the complete participation in all course units and programming tasks and the successful programming of a small final project. This, however, will be more fun than stressful for you after this intensive Deep Learning training.
The agenda is written in English due to the numerous technical terms. Descriptions as well as course material will be provided in German. You can book the course either in German or English.
All seminar contents are individually adapted to the wishes of our participants. They can vary depending on the level of knowledge and will be defined together with the seminar leader on day 1. In this Deep Learning with Python seminar you have the choice of the following topics:
Deep Learning for Python Programmers Day 1
Introduction to Machine Learning
In this short section we cover questions such as
- What is ML?
- Which Problems are suitable for ML?
- Unsupervised Learning
- supervised Learning
- semi-supervised Learning
- self-supervised learning
- ML workflow and lifecycle
Introduction to Deep Learning
- Fundamentals Deep Learning
- Stochastic Gradient Descent
- Activation Functions
- Cost Functions
- Regression and Classification in PyTorch
Deep Learning for Python Programmers Day 2
Overview of the PyTorch Ecosystem
Presentation of some important projects from the PyTorch ecosystem. Depending on the interest of the participants some of the projects can be can be discussed in more detail. (Please provide feedback on projects for which this is is desired two or three weeks before the start of the course).
- Models (torch.modelzoo)
- Albumentations (Image Augmentation)
- Huggingface (Transformers, NLP)
- MMF for multimodal applications
- ParlAI for dialog models
- PyTorch Lightning
- PyTorch 3D for 3D computer vision
- PyTorch Geometric for deep learning on graphs
Pre-trained models:Transormer, VGG, MobileNet
- Attention and Transformer models
- Self-supervised training strategies
- Fine-tuning of pre-trained networks
- NLP with Huggingface Transformers
- Transfer Learning for Computer Vision with fast.ai
- Object Recognition with TorchVision
Deep Learning for Python Programmers Day 3
Neural Networks for different data types
- Image Processing and Convolution
- What is convolution? What are CNNs?
- Applications of CNNs to image processing
- Applications of CNNs to Natural Language Processing (NLP)
- Recurrent Neural Networks
- Sequential data, time series
- Autoregression for time series models
- Recurrent neural networks (RNNs)
- Prediction of time series with RNNs
- Long Short Term Memory Networks (LSTMs)
- What are LSTMs?
- Text classification with LSTMs
- Problems of LSTMs
- Long Distance Problem
- efficient implementation
- GANs: Generative Adversarial Networks
- How do GANs work?
- Training a GAN
- Generating realistic images with GANs
- Learning interpretable semantic representations with Info GANs.
- Debugging of models
- Optimization of models