Proposed Dataset for Lesson 84. Creating a Straight Line Dataset to See if Our Model is Learning Anything

[ChrisLeeSG]

First of all, I wish you thank you for putting together this course. I've always wanted to learn neural network and realizing my dream through this course.

84. Creating a Straight Line Dataset to See if Our Model is Learning Anything

Idea / Proposal

Use a linear classification dataset so that we can truly test if our model is capable of performing binary classification.

Benefit

Using this dataset, we don't have to make any changes to our model, loss function and optimizer when we are testing if our model can learn anything.

How

Create a straight line and classify randomly generated points above the line as 1.0 and points below the line as 0.0.

My Code

import torch 
from torch import nn

# Assign weight and bias for a straight line
weight = 0.8
bias = 0.1

# Randomly assign X1 and X2 values
X = torch.rand(1000, 2)

# Calculate line using the equation: line = weight*X1 + bias
line = weight*X[:, 0] + bias

# For y, assign  0 if X2 is below line and 1 otherwise
y = X[:, 1] - line
y = torch.where(y < 0, torch.tensor(0.0), torch.tensor(1.0))

Result

Using this dataset to train and test our model, proves that it can perform a binary classification on this dataset.