This deep learning training teaches how neural networks are built, how they are trained and how to spot poor models, and it is aimed at developers and analysts with Python and statistics foundations. You implement small networks in PyTorch, watch loss curves, apply regularisation and compare models using clean validation. The seminar is designed for intermediate learners. Scope and data examples are agreed with you when you enquire.
What you’ll take away.
- Explain the structure of a neural network with layers, activation functions and loss function
- Follow the forward and backward pass and gradient descent on an example
- Implement networks in PyTorch with Dataset, DataLoader and a training loop
- Spot overfitting in learning curves and counter it with dropout, weight decay and early stopping
- Compare models with training, validation and test data and suitable metrics
Your training in detail
- Who is it for?
- Software developers, data analysts and aspiring ML engineers who want to go beyond classical machine learning.
- Prerequisites
- Python skills, foundations in statistics and linear algebra, and first experience with NumPy or pandas.
- Format & planning
- Suggested format: 3 days live online or on site, split into modules. Scope depends on prior knowledge.
Agenda: from understanding to application
Our independently developed agenda proposal. Focus areas, lab environment and scope are agreed before booking.
From the perceptron to the multilayer network with neurons, weights, bias and activations such as ReLU and softmax.
Learning through loss functions, backpropagation, gradient descent and optimisers such as Adam.
PyTorch in practice with tensors, autograd, nn.Module, DataLoader and the training loop.
Overfitting and regularisation with learning curves, dropout, weight decay, data augmentation and early stopping.
Architectures at a glance: convolutional networks for images, recurrent networks and transformers for sequences.
Fair model comparison with data splits, cross-validation, metrics, fixed seeds and a results log.
Suggested exercise: On the FashionMNIST dataset you train two networks of different size, compare their learning curves, curb overfitting in the larger one and justify your model choice in writing.
Exams, licences & credentials
There is no single authoritative external exam for deep learning. Attending alone awards no vendor certificate. Exam fees and licences are only included if the offer confirms them.
Vendor references & review date
Sources checked on 2026-11-03. These sources describe the technology or external exam requirements. They do not imply a partnership with NextNowa.
Your learning content
These objectives describe the training topic. The specific syllabus, duration, dates, trainer and price are agreed before booking.
Frequently asked questions
What maths do I need for deep learning?
Helpful are foundations in linear algebra (vectors, matrices), calculus (derivatives) and statistics. You do not need to prove anything from memory, but gradients and probabilities should feel familiar. The training explains the ideas with small worked examples and PyTorch code rather than long derivations.
How do I recognise overfitting in a neural network?
Training error keeps falling while validation error stalls or rises again. Learning curves show this gap clearly. Remedies are more data, dropout, weight decay, augmentation, a smaller model or early stopping. It is important that the test data stay untouched in the process.
PyTorch or TensorFlow: which should I learn?
Both are widespread, and the concepts transfer easily. PyTorch is often the first choice in research and many open-source projects, while TensorFlow is common in existing production systems. We work with PyTorch so you learn the concepts through code. You can tell us which framework your company uses when you enquire.
Does Deep learning foundations include a certificate?
There is no single authoritative external exam for deep learning. Attending alone awards no vendor certificate. Exam fees and licences are only included if the offer confirms them.