Why bidirectional RNN is better than unidirectional RNN: A theoretical proof
deep learning, theoretical, recurrent neural networks DeepwizAI deep learning, theoretical, recurrent neural networks DeepwizAI

Why bidirectional RNN is better than unidirectional RNN: A theoretical proof

Ever wondered how Machine Learning research papers come up with their own mathematical proofs? In this article, we will provide you with a step-by-step, theoretical proof of why a bidirectional Recurrent Neural Network (RNN) performs empirically better than a unidirectional RNN. Our proof is novel and does not exist in Deep Learning literature. By going through this article, you will learn how to write your own proofs!

Read More
Why does regularizing the bias lead to underfitting in neural networks?
deep learning, machine learning, regularization, bias, regression, generalizability DeepwizAI deep learning, machine learning, regularization, bias, regression, generalizability DeepwizAI

Why does regularizing the bias lead to underfitting in neural networks?

This article explains the reason behind the taboo of not including the bias parameter in regularization while also explaining the substantial role it plays in algorithms such as linear regression and neural networks. So read on!

Read More
Why Random Shuffling improves Generalizability of Neural Nets
machine learning, deep learning, generalizability, gradient descent DeepwizAI machine learning, deep learning, generalizability, gradient descent DeepwizAI

Why Random Shuffling improves Generalizability of Neural Nets

You must have both heard and observed that randomly shuffling your data improves a neural network's performance and its generalization capabilities. But what is the possible reason behind this phenomenon? In this blog, we provide an intuitive explanation for the same.

Read More