AI & Deep Learning Course Training in USA

This module extends the concepts introduced in the previous chapters to multiple hidden layers.

Convolutional neural networks are a vast improvement on the MLP algorithm which has greatly increased the efficiency of ANNs. We discuss the fundamental concepts and basic building blocks of CNNs.

Images data is quite different from the regular tabular data and needs to be handled in a particular manner. Since machine learning algorithms rely on mathematical computations, images need to be processed and converted into numerical (preferably vectorised) representations. This module addresses how that is accomplished.

RNNs are a particular type of deep learning architecture that has had a very high success rate in image and object detection problems. This module introduces the architecture of RNNs and the different variations.

LSTMs are particular types of RNNs. They have something called a ‘memory’ cell which ‘remembers’ information as it flows through the network. Hence this makes LSTMs best for forecasting type problems or any use case that has a temporal component.

This architecture is another specialized type of LSTMs. This module goes into the most efficient implementation of a GRU.

Auto encoders are a type of unsupervised networks. RBMs are no longer used, but are important for historical purposes.

This type of architecture belongs to the generative class of neural networks. The differentiating feature of this architecture is that there is no connection between the cells/neurons in a single layer.

Another type of generative architecture, this class of networks have two competing networks. One network tries to create a clone of the original input while the other tries to spot the fake. After enough training the networks produce near identical clones of the original input (typically images)

This is different from Supervised and unsupervised learning, which uses a concept called reinforcement. It builds upon the Markov Decision Process and builds an architecture that differs from both supervised and unsupervised approaches.

This is the part where you use all your knowledge to build a chatbot from scratch. We start from a simple rules based one, to a more complex one using NLP techniques.