Supervised learning is a type of machine learning where the model is trained on a labeled dataset, meaning each training example is paired with an output label.
In other words, supervised learning teaches a machine by example:
- You give it inputs and the correct answers, so it learns to predict the answer for new, unseen inputs.
Common types of supervised learning
As it relates to email security, a common type of supervised learning is spam detection.
For example:
- Goal: Classify incoming email as either spam or ham (not spam)
- Input: Emails labeled spam or ham
- Output: For each new email passing through the supervised learning algorithm, label it spam or ham
How does it differ from unsupervised learning?
Unsupervised learning involves training a model on unlabeled data.
In other words, the machine learning algorithm tries to find patterns or structures in data without any labels.
For example:
- Goal: Segment a customer base based on their behaviors
- Input: Customer data
- Output: Groups of similar customers
How does supervised learning differ from reinforcement learning?
A core principle of reinforcement learning is that it interacts with an environment and gathers feedback via rewards or penalties.
For example:
- Goal: Learn how to play chess
- Input: At every step, the algorithm receives input from the environment in the form of a state. The state captures the current situation of the environment. In this example, the state is the layout of the chess board.
- Output: The algorithm then produces an output—an action taken based on the current state. For each action’s outcome, a reward signal is generated, which can be either positive or negative (i.e., a penalty)
Conclusion
On top of everything else, Founders are responsible for spotting trends and taking action.
Continual improvement puts one in a position to learn and win. This is the goal of the Founder's guide to AI fundamentals.
See related: Scaling laws: The Founder's Guide to AI Fundamentals