I finished the “Deep Learning Onramp” introductory course held on the Matlab Training platform.


The course provides the basis for dealing with Deep Learning in Matlab. Some theoretical points are addressed such as the structure of CNN networks (Convolutional Neural Network) and various practical topics for implementation.

What about Deep Learning ?

Deep Learning is a Machine Learning technique that uses deep neural networks to make predictions about the content of an image. The network calculates various features and uses them for classification. It fits into the research field of machine learning and exploits representation at different levels, where at each level we look for hierarchies of characteristics, factors or concepts.

It was born in the 80s but only in recent years has it overcome some barriers that become widely used, especially thanks to the increase in the computing power of common computers and thanks to the use of GPU-based parallel computing systems.

deep learning
Typical structure of a convolutional network – Wikimedia

Among the different applications it is possible to:

  • Automatic speech recognition
  • Image recognition
  • Discovery of drugs and toxicology
  • Recommendation systems
  • Bioinformatics
  • Autonomous driving
  • Fraud Identification
  • Medical diagnoses

Machine learning algorithms

The different algorithms exploit some key points. They use non-linear levels in cascade which perform the tasks of extracting indicators and transforming them. The output of each level will be input to the next level. The algorithms can be both supervised and unsupervised. Furthermore, multiple levels represent multiple levels of abstraction that form a hierarchy of concepts. Therefore the machine can autonomously classify the data and give them a hierarchy, often finding the best combinations to solve the problem.