4. What is Artifical Neural Network ?
Neural Network
Neurons Connections
between them
An artificial neuron network (ANN) is a computational i.e.,an
information processing model that is inspired by the way
biological nervous systems(such as the brain) process the
information.
ANNs are considered nonlinear statistical data modeling tools
where the complex relationships between inputs and outputs
are modeled or patterns are found.
7. Perceptron
A perceptron is a most fundamental unit of neural
network(Artificial Neurons) that does certain computations
to detect feature or business intelligence in the input data.
The perceptron is a linear model for supervised learning
used for Binary Classification.
Perceptron consist of 4 parts:
i. input
ii. weights & bias
iii. summation function
iv. Activation Function
v. Output
Perceptron Learning Rule
Perceptron learns the weights for the input signals in
order to draw a linear decision boundary
Two types of perceptron:
a) Single Layer Perceptron
b) Multi-layer perceptron
10. Feed-Forward Neural Network:
It is also known as
Multi-layered Neural
Network
Information only travels
forward in the neural
network, through the
input nodes then through
the hidden layers(1 or
more)and finally through
the output nodes.
Capable of handling the
non-linearly separable
data
Layer present between
input and output layer
are called HIDDEN Layers.
11. Backpropagation Algorithm:
Backpropagation, an abbreviation for
”Backward propagation of errors” is
common method of training ANN.
The method calculates the gradient of a
loss function w.r.t all the weights in the
network.
The gradient is fed to the optimizer in
order to minimize the loss function.
The Backpropagation algorithm looks for
the minimum value of the error function in
weight space using a technique called the
delta rule or gradient descent.
The backpropagation learning algorithm
can be divided into two phases
i. Forward Propagation(propagate)
ii. Backward Propagation(update weights)
12. Algorithm
Step 1:Initializtion
Randomly set all the weights threshold levels of n/w.
Step 2: Forward computing:
compute the hidden vector h on hidden layer
zj =φ(∑i vijxi)
compute the o/p vector y on o/p layer
yk= φ(∑i wjkzj)
Step 3: Calculate the Total Error
Check the difference between y(actual o/p) and ŷ (predicted
o/p)
E=1/2(y-ŷ)2
Step 4: Backward computing
Finding the derivative of the error
Calculating the partial derivative of the error w.r.t
weight
Update the weights: ∆ wj = - ɳ ∂E
∂wij
wj=wj+ ∆ wj
24. Future of ANN: Deep
Learning
Artificial Intelligence
,Machine Learning and
Deep Learning are
interconnected fields
Machine learning and Deep
learning aids AI by
providing a set of
algorithms and Neural Net
to solve a data driven
problems
25. Deep Learning
Deep learning is a subset of machine learning in artificial intelligence
(AI) that has networks capable of learning unsupervised from data
that is unstructured or unlabeled
30. Advantages & Disadvantages
of ANN
Parallel processing ability
Information is stored on an
entire network ,not just a
database
Fault tolerance means
corruption of one or more cells of
the ANN will not stop the
generation of output
Gradual Corruption means the
network will slowly degrade over
time, instead of a problem
destroying the network instantly
The lack of rules for
determining the proper network
structure
The requirement of processors
with parallel processing abilities
makes ANN hardware dependent
The lack of explanation behind
probing solutions
Generation of lack of trust in
the network
Advantages
Disadvantages