Deep Learning - Chorale.txt - Deep Learning Deep Learning...

Deep Learning
Deep Learning is at the heart of many cutting edge Artificial Intelligence (AI)
applications. In this course and the successive ones, you will understand
various aspects of deep learning that will help you to
Learn Deep Learning Algorithms
Architecture and
Applications.
Neural Networks is a set of algorithms designed to learn the way our brain
works.
The biological neurons inspire the structure and the function of the neural
networks.
Each artificial neuron is modeled as follows:
Each neuron receives inputs.
Adds weights and biases to the input.
Sum inputs with weights and bias.
This triggers the Activation function.
Activation function will take the weighted sum of the previous step and fire the
output.
Supervised Learning
In Supervised Learning, you give the list of Inputs and Outputs for a Neural
Network to Learn.
Based on the actual output, the Neuron adjusts its weights and biases. This is
achieved by a process called Training.
Neural Networks learn using Back Propagation.
When the input is combined with the weights and bias to trigger the activation,
it is called forward propagation.
When the error estimates are propagated backward to adjust the weights and
biases, it is called backpropagation.
Shallow Learning - Tasks and Algorithms
Some of the tasks that can be grouped under shallow learning are:
Feature Extraction - Preprocessing
Mapping specific features into vector space - Kernel methods
Rule-based decision making - Decision Trees
Combining various estimates - Ensemble methods
Deep Learning
In deep learning, you have two or more hidden layers. This allows the algorithm
to perform multiple tasks.