Machine Learning — techniques & algorithm types

Source: Deep Learning on Medium

So what is Machine Learning?

Machine learning is a technique with the goal to achieve a type of algorithm with the goal to teach computers to led by example. This simply means it enables computers to self-learn without being continuously programmed. It’s the computer’s job to identify data patterns and improve the algorithms and sort out future data on how it will be sorted. For those who have taken statistics, it’s very common to notice that machine learning is similar to statistics and it can be hard to differ from the two.

Why is machine learning important?

Given that in our world we generate data everyday — making it harder for humans to make sense of the data. Here is where automation comes in — humans learning from the data through machine learning and understanding the data through automation.

Types of machine learning algorithms

Algorithms = series of steps the computer follows to solve a problem.

  • Reinforcement Learning : In this type of learning the computer learns to behave rationally, the key here is to maximize reward. The key here is learning through trail and error. An example of reinforcement learning, can be seen with IBM’s Watson on Jeopardy where the computer answered questions and how to wage money through the daily doubles.
  • Unsupervised Learning : This type of learning is used when the data is unlabeled. An example, predicting that all yellow flowers are sunflowers/all animals with four legs are cats. Here the computer notes that all cats belong in one cluster and notes that other four legged items/animals don’t belong in this class. The key here is that Unsupervised learning is all about description/summarization — the machine learns best through description.
  • Semi-Supervised Learning: This is the combination of Supervised and Unsupervised using labeled and unlabed data. The goal here is for the machine to learn to label unlabed data and improve classification performance.
  • Supervised Learning : This type of learning is used when the data is labeled. It best to see Supervised Learning about approximation and a generalization. Approximation because the data is correctly labeled and it tells you there and then that this is what it should be.

So what what about the data?

Data is key in machine learning. It is important to think of data and the algorithms as equally important to each other. Data with the right touch can surely bring on a well output.