Hey Everyone, I am Paras really excited to share you the application of Convolutional Neural Networks and the resources, where you can learn and understand them. So let’s get started…
CNN’s achieve state of the art results in the variety of problem areas including Voice User Interfaces, Natural Language Processing, and Computer Vision.
Voice User Interfaces – WaveNet Model
In Voice User Interfaces, Google made use of CNN’s in its recently released WaveNet model. WaveNet takes any piece of text as an input and does an excellent job of returning computer-generated audio of a human reading text. If we supply enough samples of your voice to WaveNet model it is possible to train it to sound just like you.
Read about the WaveNet model.
CNN’s can be used to extract the information from the sentences. This information can be used to classify sentiment. For Ex: If the writer is happy or sad ?, or Are the movies good or not?
Learn about CNN’s for text classification.
Read about Facebook’s novel CNN approach for language translation that achieves state-of-the-art accuracy at nine times the speed of RNN models.
CNN’s and Reinforcement Learning
CNN’s can be used to train an AI agent to play video games such as Atari Breakout. These CNN based models can be learned to play games without being given any prior knowledge of what a ball is and without ever being told precisely what the controls do. The agent only sees the screen and its score but it does have access to all of the controls that you’d give to a human user. With limited knowledge, CNN’s can extract crucial information that allows them to develop a useful strategy.
If you would like to play around with some beginner code (for deep reinforcement learning), you’re encouraged to check out Andrej Karpathy’s post.
Read more about AlphaGo. Google’s Deepmind researchers used CNN’s to train an AI agent to beat human professional GO Players.
Google’s Street Maps
Google has built a better more accurate street maps of the world by training an algorithm than can read house numbers signs from the street view images.
CNN’s achieved state of the performance in all of these problem domains and there are many more of them.
If you’re excited about using CNNs in self-driving cars, you’re encouraged to check out:
- Self-Driving Car Engineer Nanodegree, where it classify signs in the German Traffic Sign dataset in this project.
- Machine Learning Engineer Nanodegree, where it classify house numbers from the Street View House Numbers dataset in this project.
- this series of blog posts that details how to train a CNN in Python to produce a self-driving A.I. to play Grand Theft Auto V.
Thanks! Stay Connected!