W3_Understanding Training data_Pattern of a Flower


I am confused as to what the purpose of the coding assignment is for week 3. We are given a dataset in the form of a flower and we have to predict what exactly? The description did not give me a good understanding and I would appreciate it if someone could help me out

The training data are arranged in the pattern of a flower (I’ll explain more about that later), but the task has nothing to do with flowers. For this problem we are given input 2d points (x1, x2) and a color associated with each point (0=red, 1=blue). The task is train a neural network that predicts color based on a 2d coordinate.

The flower shaped arrangement of points in the training data cannot be easily classified using a single decision boundary. The example in section 3 - Simple Logistic Regression is intended to demonstrate the shortcoming of using logistic regression to predict a complicated function like this.

On the other hand, a neural network containing a single hidden layer is able to learn more complex functions than simple logistic regression. The plots in section 6 - Tuning hidden layer size demonstrate the ability of the network to predict color for this complicated flower-shaped arrangement of points.

The tensorflow playground (https://playground.tensorflow.org/) is also a good demonstration of the same lesson. In this tool you pick a dataset (some of the datasets cannot be classified using a single decision boundary) and the properties of a network (number of layers, activation function, learning rate, etc…) in order to build an intuition for how this affects model accuracy and learning speed.

Ah so we are trying to predict the flower shaped data specifically or just any pattern we train it to?

The model predicts the pattern it is trained on.

Exactly. They give you some other sample datasets to try the algorithm with if you read through to the end of the notebook.

Awesome thank you so much for helping me understand!