What is Deep Learning?

What is deep learning?

Deep learning is a subset of Machine learning. It is a discipline dedicated to the study and development of machines that can learn.

While there are so many Machine learning algorithms, why should one gravitate towards deep learning algorithms? What does the learning paradigm offer that traditional Machine learning algorithms do not? Let’s explore what makes deep learning special.

Machine learning vs. deep learning

There are two key reasons that make deep learning more special than machine learning:

1. Decision Boundary
2. Feature Engineering

Decision boundary

In the case of classification problems, the algorithm learns the mapping function that separates the two classes. This is known as a Decision boundary. A Decision boundary helps determine whether a given data point belongs to a positive or negative class.

For example, imagine there is an image with stars and squares. You are asked to create a boundary to separate the two classes. The choice of the model would depend on the data distribution. Let’s visualize this:

Feature engineering

Feature engineering is the most important step when building the model for a prediction task. It involves feature extraction and feature selection. In feature extraction, all the required features for the given problem are extracted. In feature selection, all important features are selected so that it can improve the performance of the model.

Consider an image classification task. By manually extracting features from an image, it involves a solid knowledge of both the subject and the domain. But this method is time-consuming. We can automate the feature engineering process using deep learning!

Takeaway

• Deep learning is a subset of Machine learning.
• Deep learning models can learn the complex Decision boundary that separates the two classes.
• Deep learning can automate the process of feature engineering.

---