In traditional programming, a person writes explicit rules: given this input, do that. Machine learning flips that around — instead of writing the rules, you show the system examples, and it learns the rules on its own.

Once a model exists, using it looks like this:

The overall flow of a machine learning system is: data goes in, training produces a model, and the model then makes predictions on new data.

Training

The process of adjusting a model's parameters so its predictions get closer to correct, using example data.

Inference

Using an already-trained model to produce a prediction on new data it has not seen before.

Features

The input variables a model uses to make a prediction — for example, a house’s size, age, and location when predicting its price.

Labels

The correct answer associated with a training example — the actual sale price, in the house example above.

Model

The trained structure that maps features to a prediction, based on patterns learned from training data.

• Supervised learning — the model trains on examples that already have the correct answer attached (labeled data), such as emails marked spam or not spam.
• Unsupervised learning — the model finds structure in data that has no labels, such as grouping customers into segments based on behavior.
• Reinforcement learning — the model learns by taking actions in an environment and getting a reward or penalty as feedback, adjusting its behavior over time.

A minimal supervised learning example, using a small, well-known library for illustration:

# Illustrative example using scikit-learn
from sklearn.linear_model import LinearRegression

# Features: [square_feet], Labels: price
X_train = [[1000], [1500], [2000], [2500]]
y_train = [200000, 275000, 340000, 410000]

model = LinearRegression()
model.fit(X_train, y_train)  # training

predicted_price = model.predict([[1800]])  # inference
print(predicted_price)

The model isn't given a formula for house prices — it learns the relationship between square footage and price from the training examples, then applies that learned relationship to a new input.

• Confusing training with inference

  Training happens once (and periodically after); inference happens every time the model is used to make a prediction.

• Assuming more data always means a better model

  Data quality, relevant features, and how well the data represents real-world cases usually matter more than raw volume.

• Treating a model’s output as ground truth

  A prediction is a statistical estimate based on patterns in training data, not a guaranteed correct answer.