top of page

## Introductory Vocabulary and Concepts

1. Machine learning involves an input and output. The input can be one to multiple factors that mostly should be in numerical values. The output involves what one actually desires.
2. It is easiest to show and understand the predicted output with a scatterplot of this data.  Visualizing it by plotting data and looking at relationships or regressions will help understanding.

1. ​This indicates linear regression due to the linear relationship between the X and Y (input and output) and one can plot a line of best fit to predict further relationships. Where this line intersects the Y - axis can be seen as a bias. Sometimes these relationships will be linear and the line of best fit will have a linear equation, called linear regression. Some of these relationships, on the other hand, may not be linear. For example, it could be exponential and this will be modeled by a different types of regression.

More Vocabulary:

1. Types of machine learning

1. Supervised - Includes a cluster of data that has a previous actual output.

2. Unsupervised - Training data without a previous actual output. In this situation the algorithm has no ideal output to begin with and may be more unpredictable. This website will involve mostly supervised learning.

3. Semi-supervised - Some outputs are the ones we would like

4. Reinforcement learning - Training a network towards a certain goal

2. Classification - Identifying categories to which something belongs

3. Regression - Relation between output and input

4. Probability estimation - Classification based on probability

5. Weights - Strength of connection of neurons. A higher magnitude of weights indicates a stronger connection. In predicting the price of a house, the amount of bedrooms/bathrooms might be a stronger indicator

6. Bias - An assumption made by a model to make the target function easier to approximate. Like the Y

7. Slope - Slope of the line of best fit that models the relationship between input and output

8. Features - variables in the problem set that have high enough weights to build an accurate predictive model

9. Label - What we are predicting

10. Linear Regression - when the relationship between the input and output is linear and can be predicted in a linear like fashion

11. Logistic regression - Model the probability of an event, easier to implement, efficient, and easy to train.

12. Cost - Difference between an predicted and actual

bottom of page