Machine Learning (ML) is the process by which a computer looks at a large amount of data and “learns” an internal model that captures the patterns hidden inside. Now this data can be anything, it could be images, texts or more complex data like brain signals and weather forecasts.
ML is broadly categorized into 3 broad categories:
The name says it all: learning happens under supervision. The supervisor is the label attached to each data point.
<aside> 🔷
What is a label?
Data can be collected in two forms:
Think of a playlist:
So, this “algorithm” trains of labelled data and try to predict the output for never seen before data. Depending on the output types, model can perform 2 main tasks, classification and regression:
Classification: Here the model is expected to classify the unseen data into classes or groups.
For Example: Classifying a mail into spam or non-spam folder, Classifying the cell into healthy or not healthy, Classifying the given animal image into herbivorous, carnivorous or omnivores, etc.
Regression: Here our model is expected to output a value between a continuous range of values. The most famous beginner example of this is housing price prediction. Here model is expected to predict the expected housing price (which could be anything from 0 to infinity, theoretically at least) given the location, number of bedrooms, area, etc. as its input parameters
Popular supervised “algorithms” include Linear Regression, Logistic Regression, Decision Trees, Support Vector Machines, and Neural Networks. We’ll dive into each in later sections.
<aside> 🔷
Algorithms?
we don’t hard-code a rigid set of rules; instead, we design a flexible model that can “absorb” patterns from the data and generalize to new, unseen points.
</aside>