Introduction to Neural Networks, from scratch for practical learning (Part 1) Quiz

Explore fundamental neural network concepts, key data preparations, and the first steps in implementing artificial neurons practically for hands-on beginners.

1. Concept of Artificial Neuron

   Which best describes an artificial neuron in a neural network?

   1. A programming language for machine learning
   2. A mathematical function that combines inputs and generates an output
   3. A storage device for large datasets
   4. A visualization tool for data plotting

   Explanation: An artificial neuron processes multiple inputs, combines them mathematically (often using weights and an activation function), and produces an output. It is not a storage device, which refers to hardware for data storage. While neurons can be visualized, they are not simply tools for plotting. A neuron is also not a programming language; it's a basic building block in a network.

2. Importance of Data Preparation

   Why is it important to remove irrelevant or redundant columns, such as 'Unique_ID', from a dataset before training a neural network?

   1. To avoid syntax errors in code execution
   2. To make the dataset visually smaller on screen
   3. To prevent misleading patterns that don't help prediction
   4. To create more classes for classification

   Explanation: Removing irrelevant columns ensures the model learns from meaningful information rather than random identifiers. Visual compactness is not the primary reason, as machine learning models do not rely on visual presentation. Syntax errors are typically unrelated to extra columns, and more columns do not inherently create more classes.

3. Feature Selection in Neural Networks

   If a dataset contains 'Age', 'Salary', and 'Gender', why might one choose only 'Age' and 'Salary' as features?

   1. Because models cannot process text data
   2. Because 'Gender' is always irrelevant to purchases
   3. Because these features are numeric and easier to use for plotting and analysis
   4. Because having fewer columns guarantees better accuracy

   Explanation: Numeric features like 'Age' and 'Salary' are straightforward to analyze and visualize, simplifying the first steps in practical learning. While text data such as 'Gender' can be processed, it often requires additional steps, not making it unusable. 'Gender' can be relevant depending on the problem. Fewer columns do not always mean improved accuracy; relevance of features is key.

4. Goal of a Toy Classification Problem

   In a basic neural network project, what is a typical goal when using features such as 'Age' and 'Salary' to predict 'Purchased'?

   1. To determine the gender of customers
   2. To calculate total sales revenue each year
   3. To predict an individual's exact income
   4. To classify whether a purchase is likely based on given characteristics

   Explanation: The goal is to use 'Age' and 'Salary' to predict the likelihood of a purchase, a classic classification task. Calculating sales revenue or predicting income are regression or aggregation problems. Determining gender is unrelated when that feature is dropped.

5. First Steps Before Neural Network Modeling

   What is a crucial first step after loading a dataset, before building a neural network model?

   1. Selecting a random algorithm without concern for data type
   2. Uploading the dataset to a public repository
   3. Directly running training code without inspection
   4. Checking and exploring data to understand patterns and distributions

   Explanation: Examining the data helps identify issues and underlying patterns, which guides appropriate preprocessing and model selection. Skipping inspection risks errors. Randomly choosing an algorithm ignores the importance of data properties. Uploading data to a public repository is unrelated to the modeling process.