Introduction to Neural Networks

Neural networks are at the core of the majority of deep learning applications. From computer vision use cases like facial recognition and object detection, to Natural Language Processing (NLP) tasks like writing essays and building human-like chatbots, neural networks are ubiquitous.

The different offshoots of neural networks, such as convolutional neural networks (CNN), recurrent neural networks (RNNs), and Long Short Term Memory (LSTM), are a must-know.

Neural networks are the present and the future - you should start learning right now!

The above deep learning applications must have given you a fairly good idea of how useful neural networks are. They are incredibly powerful and organizations around the world are tapping into neural networks like CNN and RNN to help them find solutions to pressing problems.

Here are a few key reasons why neural networks and deep learning have become mainstream:
● Cheaper computation cost in recent years means neural networks are accessible to a wider deep learning community
● Availability of larger datasets means we can truly harness the power of neural networks
● Neural networks have consistently shown better performance on Computer Vision, Natural Language Processing (NLP), and other key domains

Phew! There are a ton of advantages to learning neural networks in the scope of deep learning.
● Neural networks can handle a huge amount of data. This is something traditional machine learning algorithms struggle with
● Neural networks have a proven record of improving on the performance of traditional machine learning algorithms
● Neural networks can implicitly detect complex non-linear relationships between variables
● Wide range of applications! Neural networks are being used across industries, roles and functions

While neural networks like CNN and RNN are inherently awesome and are changing the deep learning field, there are still a few challenges you might face.
● Computation power: Deep neural networks like CNN and RNN can take up a lot of processing power
● Interpretability: Deep neural networks might be difficult to interpret or explain
● Overfitting: Modern neural networks have a tendency of overfitting on training data (but this can be rectified with regularization)
● Large datasets: Neural networks typically need large amount of data to train on

This is the million dollar question! What exactly should you learn about neural networks? That’s where this free course by Analytics Vidhya will help you.

We will cover topics like what is a neural network, the different parts of a neural network, applications of a neural network, among other things. We have also baked in exercises after each video lesson to test your neural network knowledge!

Excellent question! Here are three key concepts you should know about the workings of a neural network:
● What is forward propagation?
● What is backward propagation (backpropagation)?
● Role of Gradient Descent in neural networks
We will cover each of these topics in the course.

There are different types of neural networks you will come across. Here are a few key ones:
● Artificial Neural Network (ANN)
● Recurrent Neural Network (RNN)
● Convolutional Neural Network (CNN)
● Long Short Term Memory (LSTM)
We’ve seen CNN and RNN getting a lot of attention lately in the deep learning community.

Python is the preferred language for learning and building neural networks in deep learning. We’ve seen a strong preference towards Python and hence we use it in the course.

If you’re new to Python, we recommend taking our free Python course.

Neural networks open up a whole new world of projects. Once you have a hang of how neural networks work, you can get your hands on a dataset and start solving problems right away. We recommend heading over to our DataHack Platform and picking up the problem that you find the most interesting or relevant.