Scribd
Upload
6 views11 pages

00005187-Deep Learning

Uploaded by

s.chavan2396
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
6 views11 pages

00005187-Deep Learning

Uploaded by

s.chavan2396
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/ 11

DEEP

LEARNING
Presented By .

1. Shashikiran Taur

2. Shubham Gautam

3. Rupesh Bavakar
Guided by.
4. Sneha Dubal
- Pranit Sir
AJENDA
▪ BSF (Binary Step Function)

▪ ANN (Artificial Neural Network)

▪ Weight & Bias

▪ Chain Rule

▪ Optimizer
BSF (Binary Step Function)
➢ In deep learning, a binary step
function is a type of activation
function used in artificial neural
1 networks.
➢ It is a simple non-linear
0
function that maps the input to
either 0 or 1 based on a
threshold.

The binary step function was one of the earliest activation functions
used in artificial neural networks, dating back to the 1950s and 1960s.
Student Criterion (OR GATE)
Threshold >= 1

▪ OR & AND represent linearly


separable data.

Student Criterion (AND GATE)


▪ XOR represent non-linearly
Threshold >1 separable data.

▪ It is only capable of learning


linearly separable data. In other
words, it can only classify data
that can be separated by a
Student Criterion (XOR GATE) straight line. However, this
limitation can be overcome by
using more advanced types
of neural networks.
ANN (Artificial Neural Network)
Input Layer:
As the name suggests, it accepts inputs in
several different formats provided by the
programmer.
Forward Propagation

Hidden Layer:
The hidden layer presents in-between
input and output layers. It performs all
the calculations to find hidden features
and patterns.

Output Layer:
Backward Propagation
The input goes through a series of
transformations using the hidden layer,
which finally results in output that is
conveyed using this layer.
Weights & Bias

Weights:
Weight are values that are assigned to the connections between
neurons in the network. Each connection between two neurons has
a weight associated with it that determines the strength of the
connection.
Bias:
Bias is used for shifting the activation function towards left or right,
you can compare this to y-intercept in the line equation.
Chain Rule Matrix 1 Matrix 2 Matrix 3

x1 F11 W11 W12 W13 W11 W12 W11


011 W21 W22 W21
W21 W22 W23
F21
021 W31 W32 W33 W31 W32
x2
F12 W11 3 F31 W41 W42 W43
031
012
F13
Y
x3
013
F22
022
x4
W new = W old – α δL / δw old

W11 3new = W11 3old – α δL / δW11 3old


Optimizer
An optimizer is an algorithm used to adjust the weights and biases of a neural network
during the training process, with the goal of minimizing the loss function.

❖ Gradient Descent:
This is a simple optimization algorithm that adjusts the weights and
biases of the network by moving in the opposite direction of the
gradient of the loss function.

Convergence Zone
Backward
Forward Propagation Propagation

DATA NN Y^ LOSS W/B


(Random Weight
& Bias)

When to Stop:
1. Early Stoppage:
epoch = 10
epoch = 5

2. Whenever you new gradients are equal to old gradients then


algorithm stop.

W new = W old
Disadvantages Of Gradient Descent
▪ Weight Updation Takes Time.
▪ Computationally Expensive.
▪ Takes Time to converge.

Advantages Of Gradient Descent


▪ Optimal Solution is Guaranteed.
▪ Smooth Curvature.

You might also like