Machine Learning Building Blocks

This Fundamental Blocks of Machine Learning Activity will help you:

• Understand a few of the most common machine learning (ML) ‘blocks’ or functions as they are used in ML applications.

• Understand the general layout or order of common ML models.

• Explain what each fundamental block of ML is and understand how to use these blocks in creating models. 

Required Materials/Resources:

1. Access to a computer and large screen (if you want share with others)

2. A Google account to access Google Colab

3. A pen or pencil

Keywords:

Machine Learning Model - A set of algorithms, or programs, which can be trained and used to detect patterns in data in order to make predictions on future data. 

Algorithm - A set of defined instructions for a computer to perform towards learning from a set of data, making predictions on a set of data, etc. In general, algorithms are the fundamental blocks of most machine learning models.

Batch - A subset of the training dataset used in each iteration for updating the weights between neurons. Batches are beneficial in splitting large datasets into more manageable pieces. 

Epoch - A complete cycle of training a machine learning model on the entire training dataset. Completing an epoch means that the model has processed all of the data points within all of the batches in the training dataset. 

Label - The known and correct output for an input data point, provided in the dataset in supervised machine learning models. The label provides a comparable outcome for the model to determine its performance from batch to batch and epoch to epoch.

Nodes/Neurons - A fundamental part of neural network machine learning models. A neuron is a mathematical formula that takes data as an input, performs an operation on the data, and provides an output to the next layer in the model. 

Loss Function - A mathematical function which provides the numerical difference between a model’s predicted output and the labeled output, commonly considered as the model's error. A loss function is sometimes also known as the cost or objective function.

Activation Function - A mathematical function which provides a logical representation of whether a neuron’s output should be activated or not based on the input data provided. The activation function allows for machine learning models to learn from nonlinear datasets.