Today I started doing a Machine Learning course on Coursera by DeepLearning.AI Stanford Online Called: Supervised vs. Unsupervised Machine Learning. It has been incredible how comprehensively the topics have been introduced in this course. Let me give a brief summary of what I learned.
Machines usually learn in two different types of ways: supervised and unsupervised. Supervised learning refers to training a machine using input and output pairs. The machine is thus, able to make ‘connections’ between these points and then use this connection to predict outputs for unseen data. Unsupervised Learning refers to training a machine using data that has only inputs, the machine therefore, has to look for trends in the given data set and then use these trends to make predictions about unseen data.
One major Machine Learning technique is called Regression, where a machine must make a estimate that can take any value over an infinite range. Given this lets try and put what we have learned to explain a regression model. Let’s say we have inputs which we call features and they each come with their corresponding output, in other words, we are dealing with a Supervised Learning situation. We chose to model the situation between the input and outputs as a mathematical function f = wx + b where f represents the output value and x the input value. The variables w and b are our parameters. since we are given the correct output values we can keep adjusting our parameters till we get a ‘connection’ (a straight line in this case) that best fits the input output pairs (one could think of this as the line that is closest to all the input output pairs when plotted on a graph. What we come up with at the end of the day is a model that we have essentially trained!
Please feel free to give me some feedback on this!