Learning from Data Flashcards
What are the three main categories of machine learning methods? Provide an example task for each.
Supervised: Predicting house prices based on size and location
Unsupervised: Grouping of customers by their spending habits
Reinforcement: Training a self-driving car
In supervised learning, what do x’s and y’s represent?
X is the input and y is the output.
What function g(z) is used in logistic regression? Why do we use this function?
It’s the sigmoid function and is used to predict the output as a probability for classification.
Why can’t linear models learn non-linear data distributions?
Because linear models are constrained to linear decisions.They can’t comprehend complex relationships like curves.
What is the key difference between classification and regression in supervised learning?
Classification has a finite set of outputs, regression has an infinite set of outputs.
Name three problems that are formulated as supervised learning tasks.
Image Recognition, Sentiment Analysis, and Imitation Learning/Behavioral Cloning
What is a decision boundary in the context of classification problems?
It’s a line or surface that separates the data into different classes.
What is 𝛼 in the gradient descent formula?
It’s the learning rate.
What makes neural networks capable of handling non-linear relationships?
They use non-linear functions, like sigmoid, to learn and approximate complex patterns.
What is the difference between linear regression and logistic regression?
Linear regression is used for linear data to get a value from an infinite set of outputs, whereas logistic regression is used for classification, to classify data into finite categories.
What is the range of output values for the sigmoid function?
0, 1.
Name three examples of unsupervised learning problems.
Dimensionality Reduction, Clustering of Data, Topic Modeling
What is the purpose of the cost function in supervised learning?
It measures the error between the predicted values and the actual values. Telling how good the machine is learning with it’s data set.
What relationship exists between model complexity and the ability to generalize to unseen data?
High complexity may overfit the data, easier to train, but will tend to memorize it’s training data.
Low complexity may underfit the data, tougher to train, but more likely to generalize better.
