Get Started with AI - Module #4 - Machine learning and prediction machines
Machine learning systems provide us with an artificially intelligent toolbox – a set of technologies that can help us solve problems in organizations, society and for people!
However, in order to truly understand the toolbox’s potential for use, we need to dive a little deeper into machine learning and what we mean when we talk about prediction machines. Further, we need to explore how those algorithms work to compute and learn from data, and how that is then taken to help with various tasks needed to solve problems in the real world.
Practice questions
So, now we should have a fair grasp of the fact that machine learning (ML) is a branch of AI and provides a set of technological tools to help us make sense of something in the world, represented by data. Further, the ML system can be used to help perform various tasks needed to solve problems or answer questions we might be having (e.g. through prediction).
From this point of view, we can see the crucial role that data plays for training ML algorithms. In the lecture we learned about four types of approaches that ML systems learn from data: supervised learning, unsupervised learning, reinforcement learning, deep learning. What is the main feature for each approachin how the system learns from data?
Take each of these approaches now, and let’s revise:
- What type of data is used for supervised learning?
- What type of data is used for unsupervised learning? What is a common technique used by systems to learn from data in this approach?
- What is the critical characteristic of reinforcement learning (i.e. how the system learns)?
- How does a deep neural network process (and learn from) data and what makes a deep learning neural network deep?
