Overview

What is machine learning?

Branch of Artificial Intelligence that focuses on algorithms and data to give computers the ability to learn without being explicitely programmed.

Difference ML vs AI

Machine learning is focused on teaching a machine how to perform a specific task and produce accurate results.

Artificial intelligence is a broader umbrella that focuses on mimicking human intelligence. Sub-fields of AI:

  • Machine learning
  • Deep learning (subset of machine learning that uses artificial neural networks)
  • Computer Vision
  • Natural Language Processing

Some Terminology

  • Model: A function that takes in an input, runs calculations and produces an output with probabilities.
  • Machine learning algorithms: Convolutional Neural Network (CNN), Naive Bayes, Long Short Term Memory (LSTM), ...
  • Weights: How much importance to give to each feature of the dataset to improve the model's accuracy to perform a certain task. These weights are adjusted as the model is trained on the data and the accuracy changes. Learned traits.
  • Overfitting: When the model fits too closely to the training dataset so it is not able to predict accurately new data.
  • Classes or Labels: A one-word description of an individual piece of data. For example, "dog", "cat" for images of pets.

Types of machine learning

Supervised learning

In supervised learning, the computer is given a labelled data set and the algorithm is going to find a correlation between the data and the labels.

Unsupervised learning

In unsupervised learning, the data is provided to the machine learning algorithm without being labelled.

Semi-supervised learning

Combines supervised and unsupervised learning. Some of the data is labelled and some isn't. Benefits? Can achieve higher accuracy and efficiency.

Reinforcement learning

In reinforcement learning, the training method is based on rewarding the desired behavior and punishing the undesired one. Learning through trial and error. More useful in games where there is an environment where you train an agent and actions can be taken in this environment.

Tools