Du är här

Course Schedule - Ai & Machine Learning

Back to course description and registration


During the course, the participants will be able to practice on all steps necessary to train, test (verify), integrate (on HW), and validate a machine learning model.

The exercises will be split over the two days. The first day covers exercise setting up the workflow and training an initial model and the second day handles testing and deployment. After completing the course, each participant will be allowed to bring home the Raspberry Pi containing their trained and tested model. 

Schedule – Day 1: Introduction of AI and Machine Learning    

During the first day of the course we will be taking an in-depth look at AI and machine learning from an advanced, professional perspective. We will learn how AI/ML can be used in practice, about its limitations and when one should avoid such approaches.

During the day, the course participants will set up a working machine learning environment and train a neural network.  

Introduction 

  • Definition of AI/ML  
  • Comparison to conventional approaches 
  • Common practices  
  • Application areas  

Traditional ML approaches  

  • Supervised, Unsupervised, Reinforcement     

Workflow  

  • Setting up the work environment (development environment)  
  • Identifying and categorizing the problem    
  • Choosing appropriate algorithms & metrics  
  • Data processing  
  • Gathering, cleaning, preprocessing, augmentation  
  • Model training  
  • General workflow/pipeline  

EXERCISES DAY 1

Connect the theoretical concepts to actual implementations.

Setting up the workflow

  • Set up the environment & pipeline 

First steps to ML deployment

  • Import and verify the data    
  • Train the neural network   

Schedule – Day 2: Data + V&V   

The second day will focus more on verification and validation of machine learning. Since the data used to train a ML model is essential (garbage in, garbage out), some time will be spent focusing on the data itself. In addition, we will discuss how to ensure that a ML-function actually accomplishes the task it is assigned to do.

The course will end with a discussion of potential risks of machine learning. During the day, the course participants will be able to test their trained model from the day prior, as well as transfer it to a hardware platform in order to deploy the model.

Data    

  • Dataset (Real-world vs Synthetic)    
  • Data quality  

Bias    

  • Data gathering  
  • Training  
  • In development and testing   

Verification and validation  

  • Quality assurance using test levels 
  • Choosing a metric  
  • Simulators  

Risks  

  • Potential problems with OpenSource  
  • Ethical AI    
  • Safety critical applications  

EXERCISES DAY 2

Connect the theoretical concepts to actual implementations.

Testing the model

  • Test the model locally (on computer)  

Deploying on HW & testing

  • Deploy the model on the Raspberry Pi
  • Test the model on the Raspberry Pi

REGISTRATION

Send an email to tomas.sjogren@infotiv.se to either sign up or ask for more details.


Back to course description and registration