What is Machine Learning?

Machine Learning - What is Machine Learning?

Consider a common problem: predicting house prices. If we plot the prices of various houses against their sizes (e.g., in square feet), we might see a scatter of data points. For instance, imagine a graph where the horizontal axis represents ‘House Size (sq. ft.)’ and the vertical axis represents ‘House Price ($)’. As house size increases, the price generally tends to increase, so the points would likely show an upward trend.

Machine Learning allows us to analyze this data. We can feed these known data points (house sizes and their corresponding prices) to a machine learning algorithm. The algorithm’s task is to find a relationship, often by trying to fit a ‘best-fit’ line through these points. Once this line (or model) is determined, we can use it to predict the price of a new house given its size. If we input a known house size from our test data, the machine will return an estimated price based on the learned line. We can then compare this prediction to the actual price to see how well our model works. This line can also be extrapolated to estimate the price of a house larger than any in our original dataset, for example, a 3000 sq. ft. house. This process of predicting a continuous value (like price) is a type of statistical task called regression. If the relationship is represented by a straight line, it’s specifically called linear regression.

However, relationships in data are not always linear. In many cases, the connection between input features (like house size) and the output (like price) might be more complex, better represented by a curve than a straight line. For example, imagine a dataset where the data points clearly follow a curved pattern, not a straight one. The task then becomes finding the ‘best-fit curve’ that accurately models this non-linear relationship, which can then be used for predictions.

Machine Learning, at its core, is about using algorithms that can learn from data to find such relationships or patterns. These algorithms often employ optimization techniques – iterative processes that adjust the model’s parameters (like the slope and intercept of our line, or the coefficients of our curve) to minimize prediction errors on the known data. By learning from past examples, the machine develops the ability to make predictions or decisions on new, unseen data.

Next, let us look at the different categories of Machine Learning.