Introduction

In this intensive workshop you will learn how to develop, train and apply deep learning models to different types of data. You will gain insight into the operation of neural networks , learn how to build a model from scratch with TensorFlow and PyTorch , and apply advanced architectures such as CNNs for image recognition and RNNs or LSTMs for time series or text.

You will discover which hyperparameters impact performance, how to prevent overfitting with regularization and dropout techniques, and how to use pretrained models and transfer learning to create value faster with less data. We will also discuss the role of loss functions, activation functions and optimization algorithms such as Adam , SGD and RMSprop .

This training is aimed at professionals who want to not only understand deep learning, but also apply it to their own data workflows — with a strong focus on hands-on experience and best practices.

Learning objectives

• Understanding the structure and operation of feedforward neural networks (layers, weights, activations, backpropagation)

• Build and train your own dense network with TensorFlow (Keras) or PyTorch

• Apply deep learning models to structured data, text and images

• Using Convolutional Neural Networks (CNNs) for image classification, object detection or feature extraction

• Using Recurrent Neural Networks (RNNs) and LSTMs for sequential data such as time series or natural language

• Apply hyperparameter tuning: learning rate, batch size, epochs, activation functions, etc.

• Selecting loss functions and evaluation metrics for classification and regression tasks

• Use of pretrained models and application of transfer learning (such as ResNet, BERT, EfficientNet)

• Working with callbacks, model checkpointing and early stopping

• Introduction to model interpretation and visualization of activations

Approach and working methods

The training consists of short theoretical blocks, alternated with extensive hands-on assignments. You work in Python with Jupyter Notebooks , and choose whether to work with TensorFlow (Keras) or PyTorch , depending on your preference or experience.

We use realistic datasets (e.g. MNIST, Fashion-MNIST, time series, NLP datasets, classification tables) and train locally or in the cloud with GPU support (e.g. via Google Colab or your own environment).

You will learn, among other things:

• Setting up and adapting architectures

• Debugging model performance

• Monitor training and evaluation with TensorBoard or similar tools

• Prepare and augment datasets

We prefer to tailor the training to your domain or project context — think of applications in finance, healthcare, industry or media.

For whom

This training is intended for AI engineers, ML specialists, data scientists and developers who want to apply deep learning to their own data and want to be able to build, adapt and improve models themselves.

A good command of Python and basic knowledge of machine learning is highly recommended.

Interested in this training?

Feel free to contact us. We are happy to help you with a customized training that matches the level and objectives of your team or organization.