Machine Learning

Step 1: Understanding the learning problems.
Classification is a supervised learning problem where the output is categorical, meaning it takes on a finite set of values, such as "sunny" or "rainy."

Step 2: Analysis of options.
- (A) Classification: Correct, classification is used when the output is categorical.
- (B) Clustering: Incorrect, clustering is an unsupervised technique used for grouping similar data points, not for categorical output.
- (C) Regression: Incorrect, regression deals with predicting continuous values, not categorical ones.
- (D) Optimization: Incorrect, optimization focuses on finding the best solution to a problem but does not specifically involve categorizing data.

Step 3: Conclusion.
The correct answer is (A) Classification.

Step 1: Understanding the learning types.
In supervised learning, the agent is given input-output pairs, where the output is labeled. The agent then learns a function that maps inputs to outputs.

Step 2: Analysis of options.
- (A) Supervised: Correct, in supervised learning, input-output pairs are used to train the model.
- (B) Unsupervised: Incorrect, in unsupervised learning, there are no output labels provided.
- (C) Reinforcement: Incorrect, reinforcement learning involves agents interacting with an environment and learning through rewards or punishments.
- (D) Semi-supervised: Incorrect, semi-supervised learning involves both labeled and unlabeled data.

Step 3: Conclusion.
The correct answer is (A) Supervised.

• Each model sees a slightly different subset of the data.
• The models are trained independently.

• For classification, predictions are combined using majority voting.
• For regression, predictions are combined using averaging.

• Bias: Error due to overly simple assumptions in the learning algorithm.
• Variance: Error caused by the model being too sensitive to the training data.

• Very well on training data
• Poorly on testing or new data

• = actual value
• = input feature
• = regression coefficients
• = error term

• All errors become positive
• Larger errors are penalized more heavily

• Sigmoid: Output range
• ReLU: Output range
• Tanh: Output range
• Softmax: Output range and used for probability distributions