Module Question 3

Machine Learning

1. Supervised Learning vs. Unsupervised Learning

Question:
What is the key difference between supervised and unsupervised learning? Provide an example for each.

Status:
Completed

Description:
The main difference between supervised and unsupervised learning is the type of data used to train the model. Supervised learning uses labeled data, which means the input data already has a known output or target. The model learns the relationship between the input and the correct output so it can make predictions on new data.

For example, a spam email detection system can use supervised learning by training a model with emails labeled as “spam” or “not spam.”

In contrast, unsupervised learning uses data without predefined labels. The model tries to discover patterns, relationships, or groups within the data by itself. For example, a company can use unsupervised learning to group customers based on their purchasing behavior without assigning categories in advance.

Evidence:
IBM – What is Supervised Learning?
https://www.ibm.com/think/topics/supervised-learning

IBM – What is Unsupervised Learning?
https://www.ibm.com/think/topics/unsupervised-learning

2. Overfitting in Machine Learning

Question:
What is “overfitting” in a machine learning model, and why is it a problem?

Status:
Completed

Description:
Overfitting occurs when a machine learning model learns the training data too closely, including unnecessary details or noise. As a result, the model may perform very well on the training data but perform poorly when it is given new or unseen data.

Overfitting is a problem because the goal of machine learning is for a model to generalize well to new data, not simply memorize the training examples. For example, a model trained to recognize cats and dogs may have high accuracy on the images used during training but struggle when it receives new images.

Overfitting can occur when a model is too complex or is trained for too long. Using more training data, regularization, and evaluating the model with separate data can help reduce the risk of overfitting.

Evidence:
IBM – What is Overfitting?
https://www.ibm.com/think/topics/overfitting

Google for Developers – Overfitting
https://developers.google.com/machine-learning/crash-course/overfitting/overfitting

3. Components of Reinforcement Learning

Question:
Describe the three main components of reinforcement learning: agent, environment, and reward.

Status:
Completed

Description:
Reinforcement learning is a type of machine learning in which an agent learns by interacting with an environment and receiving feedback from its actions.

The first component is the agent. The agent is the system or program that makes decisions and takes actions. For example, in a game, an AI player can act as the agent and decide what move to make.

The second component is the environment. The environment is the world or situation in which the agent operates. It responds to the agent’s actions and provides information about the resulting state.

The third component is the reward. A reward is feedback that tells the agent how good or bad an action was. A positive reward encourages the agent to repeat useful actions, while a negative reward or penalty discourages unwanted actions.

For example, in a robot navigation system, the robot can be the agent, the surrounding area can be the environment, and successfully reaching the destination can provide a positive reward. Through repeated interactions, the agent learns which actions are more likely to produce better results.

Evidence:
IBM – What is Reinforcement Learning?
https://www.ibm.com/think/topics/reinforcement-learning

Google for Developers – Introduction to Reinforcement Learning
https://developers.google.com/machine-learning/crash-course/intro-to-rl

4. Training Dataset vs. Testing Dataset

Question:
How does a training dataset differ from a testing dataset, and why is this separation important?

Status:
Completed

Description:
A training dataset is the data used to teach a machine learning model. During the training process, the model analyzes the data and learns patterns or relationships that can be used to make predictions.

In contrast, a testing dataset is a separate set of data that is not used to train the model. It is used after training to evaluate how well the model performs on data that it has not previously seen.

Keeping the training and testing datasets separate is important because it helps measure the model’s ability to generalize to new data. If the same data is used for both training and testing, the model may appear to perform very well because it has already seen the examples. This can hide problems such as overfitting.

For example, if a dataset contains 1,000 customer records, some records can be used to train the model while the remaining records are kept for testing. After training, the testing data can be used to determine whether the model can make accurate predictions on new customer records.

Evidence:
Google for Developers – Training and Test Sets
https://developers.google.com/machine-learning/crash-course/overfitting/dividing-datasets

IBM – What is Training Data?
https://www.ibm.com/think/topics/training-data