Start a conversation ↗

Description

Deep Learning: Understanding Neural Networks, Training and Evaluation

Build the deep learning fundamentals needed to engage effectively with data science teams

  • 0.5 days — 3.5 h
  • In-person or virtual
  • Foundation
  • Up to 6 participants

Artificial intelligence projects are multiplying across government entities and decisions are being taken using terms whose meaning remains unclear. Neural network, learning, overfitting, pre-trained model: the vocabulary circulates without everyone understanding what a model can actually learn, what it requires in terms of data and what it will never be able to do.

This short format sets out the fundamentals. Participants understand how a network learns from examples, follow the stages of training and evaluation, discover the main architectures and identify the limitations and biases to consider before launching a project.

Learning objectives

  • Position deep learning among machine learning approaches
  • Explain how a neural network learns
  • Describe the stages of training and the role of datasets
  • Interpret model evaluation metrics
  • Identify the limitations, biases and data requirements of a project

What makes this programme different

Concepts are introduced through use cases encountered in the public sector
Overfitting is observed on a worked example followed step by step
A set of questions to put to technical teams is built during the session

Programme

1From machine learning to neural networks

Understanding what depth changes

  • Supervised learning and unsupervised learning
  • The concept of the artificial neuron and the layer
  • The role of weights and activation functions
  • What depth adds compared with classical methods

2Training a model

Following the learning cycle

  • Preparing and splitting training and validation datasets
  • Loss function and progressive adjustment of weights
  • Overfitting and regularisation methods
  • Data volume and computing cost required
  • Reading learning curves

3Architectures, applications and limitations

Knowing what can reasonably be targeted

  • Convolutional networks for images and networks suited to sequences
  • Pre-trained models and adaptation to a business need
  • Evaluation metrics and interpretation of errors
  • Data bias and explainability requirements
  • Questions to ask before committing to a project

Who is it for

Public sector officers and managers, data project leaders and business profiles involved in artificial intelligence projects.

Prerequisites

No prerequisites

Dates & locations

36 scheduled dates between November 2026 and December 2027. Seats are confirmed in the order enquiries are received.

November 2026

December 2026

January 2027

February 2027

March 2027

April 2027

May 2027

June 2027

September 2027

October 2027

November 2027

December 2027

None of these dates suit you? We open additional sessions on request, and any programme can be run privately for your team.

Practical details

Before the programme
Online positioning questionnaire. Your development objectives are shared with the trainer, who tailors the practical case studies to your context.
Teaching methods
Theoretical input, workshops and practical case studies. Digital course materials and method sheets provided.
Assessment
Multiple-choice tests and role-play exercises. Assessment of learning at the start and end of the programme, with immediate and 60-day follow-up evaluations.
After the programme
One year of access to the e-learning platform. Self-assessment of the skills acquired and a 30-day follow-up session with your trainer.
How to register
Registration online or on the basis of a quotation.
Lead time
11 working days after confirmation of registration.
Accessibility
Accessible to people of determination. Contact our accessibility coordinator to design a suitable solution: contact@mpf-academy.ae
Start dates
Rolling intake: in addition to the scheduled sessions, this programme can start on request.