Problem Overview: The XOR (exclusive or) problem is a binary classification problem where the task is to learn a decision boundary that separates two classes of data. The challenge arises because XOR is not linearly separable, meaning a single straight line cannot accurately divide the two classes. This problem is often used to illustrate the limitations of simple linear models and the need for more complex models, such as neural networks, to solve non-linear classification tasks.
Truth Table :
Correct Classification: First type of results are two different cases where the regressor finds correlation between data points on the diagonals of the domain.
Challenges and Anomalies
➢ Not Converging: Random weight initialization can produce values which lead to NN not converging within max iteration threshold, thus producing this "stuck" result.
➢ Anomalies: Increasing the number of hidden layers and neurons in each layer, can push the network to converge faster but also causes NN to produce peculiar results.
Decoding the XOR Magic: Unveiling the Mechanics of Convergence
This is the gif that shows how the convergence happens in XOR classification