Improving the Model
The goal of modeling in machine learning is to ensure our model generalizes well on unseen data. Throughout our journey as data professionals who build models with neural networks, we are likely to come across two main issues: underfitting and overfitting. Underfitting is a scenario in which our model lacks the necessary complexity to capture underlying patterns in our data, while overfitting occurs when our model is too complex such that it not only learns the patterns but also picks up noise and outliers in our training data. In this case, our model performs exceptionally well on training data but fails to generalize well on unseen data. Chapter 5, Image Classification with Neural Networks, examined the science behind neural networks. Here, we will explore the art of fine-tuning neural networks to build optimally performing models for image classification. We will explore various network settings in a hands-on fashion to gain an understanding of the impact of...