Scribd
Upload
21 views24 pages

Artificial Neural Network Notes

Uploaded by

alexkunwar9817
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views24 pages

Artificial Neural Network Notes

Uploaded by

alexkunwar9817
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

Artificial Neural Networks and

Deep Learning

Artificial Neuron

Artificial Neural Networks and Deep Learning 1


An Artificial Neuron is a computational unit in artificial neural networks that
mimics the basic function of a biological neuron. It receives input signals,
processes them using weights and an activation function, and produces an
output signal.

Single Layer Perceptron


A Single Layer Perceptron (SLP) is the simplest type of artificial neural network.
It consists of a single layer of output nodes, where each node is connected to
every input. The perceptron applies a linear combination of input features and
weights, followed by an activation function (usually a step function) to produce
a binary output. SLPs are only capable of solving linearly separable problems,
meaning they can only classify data that can be separated by a straight line or
hyperplane.

Calculation (OR):

Artificial Neural Networks and Deep Learning 2


Artificial Neural Networks and Deep Learning 3
Multilayer Perceptron
A Multilayer Perceptron (MLP) is a type of artificial neural network that
consists of multiple layers of neurons: an input layer, one or more hidden
layers, and an output layer. Unlike a Single Layer Perceptron, MLPs can solve
complex, non-linearly separable problems. Each neuron in a layer is connected
to every neuron in the next layer, and the network uses activation functions
(like ReLU or sigmoid) to introduce non-linearity. MLPs are trained using
backpropagation to adjust the weights and minimize error, enabling the network
to learn from data and improve predictions.

Artificial Neural Networks and Deep Learning 4


SIGMOID Function = Activation Function

Calculation (XOR):

First:

Artificial Neural Networks and Deep Learning 5


Second:

Artificial Neural Networks and Deep Learning 6


Third:

Artificial Neural Networks and Deep Learning 7


Fourth:

ERROR (LOSS)

Artificial Neural Networks and Deep Learning 8


Absolute (error) average = 0.49
- Can be fixed by adjusting the weight so the error is as low as possible

Gradient
Gradient refers to the rate of change of a function with respect to its variables.
In neural networks, it is used to measure how much a small change in weights
affects the error, guiding the optimization process during training to minimize
the loss.

Artificial Neural Networks and Deep Learning 9


Artificial Neural Networks and Deep Learning 10
Delta Parameter
The Delta Parameter refers to the change or adjustment applied to the weights
in a neural network during the training process, based on the error gradient. It
helps update the weights to minimize the overall error or loss in the model.
First row:

Second row:

Third row:

Artificial Neural Networks and Deep Learning 11


Fourth row:

For hidden layers

First row:

Artificial Neural Networks and Deep Learning 12


Second row:

Third row:

Artificial Neural Networks and Deep Learning 13


Fourth row:

Backpropagation
Backpropagation is a training algorithm in neural networks that calculates the
error’s gradient for each weight by propagating the error backward from the
output layer to the input layer, allowing the model to adjust the weights and
minimize the error through gradient descent.

Artificial Neural Networks and Deep Learning 14


First:

Second:

Third:

Artificial Neural Networks and Deep Learning 15


Go back now:

First:

Artificial Neural Networks and Deep Learning 16


Second:

Third:

Artificial Neural Networks and Deep Learning 17


Final Step:

BIAS
Bias is an additional parameter in neural networks that helps adjust the output
along with the weighted sum of inputs. It allows the model to better fit the data
by shifting the activation function, enabling it to handle patterns that do not
pass through the origin.

Artificial Neural Networks and Deep Learning 18


MSE & RMSE

Predictions when there are more than two classes example:

Artificial Neural Networks and Deep Learning 19


Gradient Descent

Artificial Neural Networks and Deep Learning 20


Step Function

Sigmod Function

Artificial Neural Networks and Deep Learning 21


Hyperbolic Tangent Function

RELU Function (Rectified Linear Units)

Artificial Neural Networks and Deep Learning 22


SoftMax Function

Linear Function

Artificial Neural Networks and Deep Learning 23


Artificial Neural Networks and Deep Learning 24

You might also like