What is Deep Learning?

➢ Deep Learning is a subset of Machine Learning and is used to extract useful patterns from
data.
➢ The only difference between Machine Learning and Deep Learning are,
✓ Machine Learning models need human intervention to arrive at the optimal
outcome.
✓ Deep Learning models make predictions independent of human intervention.
➢ Deep learning can be classified into different types based on the architecture of the model,
such as feedforward neural networks (ANN), convolutional neural networks (CNN), recurrent
neural networks (RNN)
➢ Some of the applications of Deep Learning are...
✓ Self-Driving
✓ Gaming-Pokeman
✓ Machine Translation
✓ Computer Vision

What is Perceptron?

--->A Perceptron was a form of neural network introduced in 1958 by Frank Rosenblatt.

--->Perceptron is a Single layer neural network.

--->Its Consists of weights, the summation processor, and an activation function.

------------------------------------------------------------------------------------------------------------------------

What is Neural Network?

--->Neural Network is made up of multiple neuron.

--->The Element of neural network are Input Layer,Hidden Layer and Output Layer.

--->Each neuron is receiving input signal processing that input signal with the help of
activation function.

------------------------------------------------------------------------------------------------------------------------

What are the Element of Neural Network?

Input Layer:

--->Input layer accept input features and provides the information from outside world to neural
network.

--->No computational is performed at this layer.

Hidden Layer:

--->Hidden Layer are not exposed to outer world.

--->Hidden layer performs all sorts of computation on the features entered through input layers.
Output Layer:

--->Its bring up the information learned by the neural network to outer world.

------------------------------------------------------------------------------------------------------------------------

what is Activation Function?

--->Activation function calculates "weighted sum" of its input,add a bias and then decides whether a
neuron should

be activated or not.

--->The purpose of activation function is to introduce non-linearity into output of a neurons.

--->They are different types of Activation Function.

1.Linear Function

2.Sigmoid Function

3.Tanh Function

4.ReLu Function

5.Softmax Function

Sigmoid: This function maps any input value to a value between 0 and 1, and

is often used in the output layer of a binary classification model.

ReLU (Rectified Linear Unit): This function maps any negative input value to zero

and keeps any positive input value unchanged. This activation function is widely

used in feed-forward neural networks, especially in the convolutional neural networks (CNN) and
deep learning.

Tanh (Hyperbolic Tangent): This function maps any input value to a value between

-1 and 1 and is similar to the sigmoid function, but centers the data around zero.

Leaky ReLU: A variant of ReLU function that allows small negative values to pass

through instead of becoming zero.

Softmax: This function is used to produce a probability distribution over the output classes.

it is commonly used in the output layer of a multi-class classification model.

ELU (Exponential Linear Unit): This function is similar to ReLU but it assigns negative
values to a small negative slope.

Swish: A new activation function introduced in 2017, it is similar to sigmoid but it can be

trained more efficiently and it works better with deeper networks.

------------------------------------------------------------------------------------------------------------------------

What is Linear Function?

Equation :y = ax

--->No matter how many layers we have, if all are linear in nature, the final activation function of last
layer is

nothing but just a linear function of the input of first layer.

--->Its range from -infinity to +infinity

--->We can used linear function at one place,that is the output layer.

------------------------------------------------------------------------------------------------------------------------

What is Sigmoid Function?

Equation :A = 1/(1 + e-x)

--->It is a function which is plotted as ‘S’ shaped graph.

--->It is non linear in nature

--->Its range from 0 to 1.

--->The Sigmoid Function is used in output layer of a binary classification.

------------------------------------------------------------------------------------------------------------------------

What is Tanh Function?

Equation :-tanh(x) = 2 * sigmoid(2x) - 1

--->The activation that works almost always better than sigmoid function is Tanh function also knows
as

Tangent Hyperbolic function.

--->It is non linear in nature.

--->Its range from -1 to +1.

--->Its is used in hidden layers of a neural network as it’s values lies between -1 to 1.

------------------------------------------------------------------------------------------------------------------------

What is ReLu Function?

Equation :- A(x) = max(0,x). It gives an output x if x is positive and 0 otherwise.

--->RELU Stands for Rectified linear unit. It is the most widely used activation function.
--->Chiefly implemented in hidden layers of Neural network.

--->Its range from 0 to infinity.

--->It is non linear in nature.

--->We can easily backpropagate errors and multiple layers of neuron being activated by the ReLu
Function

--->ReLu is less computationally expensive than tanh and sigmoid function and it activate only few
neurons at a time.

------------------------------------------------------------------------------------------------------------------------

What is Softmax Function?

--->The softmax function is also a type of sigmoid function but is handy when we are trying to handle

classification problems.

--->It is non linear in nature.

OUTPUT:The softmax function is ideally used in the output layer of the classifier where we are
actually trying to

attain the probabilities to define the class of each input.

------------------------------------------------------------------------------------------------------------------------

What is Cost function?

--->The Cost function is used to measure the performance of machine learning model for a given
data.

--->Its quantifies the error between predicted values and expected values,it present results in the
form of a single

real numbers.

------------------------------------------------------------------------------------------------------------------------

Two Keys aspect with Neural Networks:

1.Activation Function

2.Cost Function

------------------------------------------------------------------------------------------------------------------------

What is Gradient Descent?

--->The Gradient Descent is an optimization algorithm that is used for minimizing the cost
function(Errors).

--->Its updates the various parameters of a machine learning model to minimize the cost function.

------------------------------------------------------------------------------------------------------------------------

What is Backpropagation?
--->Backpropagation is used to calculate the error contribution of each neuron after a batch of data is
processed.

--->Its calculates error at output and the distributes that output back throughout the network layers.

------------------------------------------------------------------------------------------------------------------------

What is DropOut?

--->Dropout is a regularization technique in with randomly selected neurons are ignored during the
training process,

that means these ignored neurons are not considered during a particular forward or backward
propagation.

--->We need to add drop out layer to prevent the overfitting.

------------------------------------------------------------------------------------------------------------------------

what is Optimizer?

--->Optimizer updates debate parameter to minimize the loss function.

------------------------------------------------------------------------------------------------------------------------

What is Epochs?

▪ One epoch consists of one full cycle of training data.

▪ Preferred value of number of training epochs is set to 1000.

--->An Epoch refers to one complete pass through the entire dataset during the training process,

in which the model updates its parameters to minimize the loss function.

------------------------------------------------------------------------------------------------------------------------

What is Deep Neural Network?

--->When a neural network contains more than one hidden layer is called Deep Neural Network.