Autonomy and AI
Originally published December 11, 2020 on my earlier blog; republished as written.
When looking at machine learning, one of the first concepts that are taught is the types of algorithms that are used to sift through the data. Understanding these types of algorithms is essential in two ways: first, you must understand how the algorithms work, second, you must understand which algorithm to use in each situation. Data comes in different forms, and the correct algorithm will determine if you have the output you are looking for.
## Supervised Learning Supervised learning is done by providing the algorithms examples of the correct data or solution to your problem. By providing the data to the system, it can use the data to train itself on examples of what you are looking for. By doing this, you teach the system to understand what right looks like.
## Unsupervised Learning Unsupervised learning looks through the data; however, it does not know which portions of the data are similar ahead of time. The system searches through the data and determines which areas are similar and groups them based on relevant information that it finds. By doing this, it outputs groups that have similar attributes or information without any knowledge of what it looked like beforehand.
## Semisupervised Learning Semisupervised learning takes data that is partially labeled. The algorithm should recognize data that is similar and rely upon you to fill in the blanks where there is missing data.
## Reinforcement Learning Reinforcement learning is the type of algorithm that uses a reward type system to determine if actions or decisions made were the correct or more correct decision. This type of algorithm can be seen in video games where programmers are using reinforcement learning to have AI play a video game without actually teaching it how to play the game. The AI will get rewards every time it makes decisions that correct move or take action that moves it through the game.