Scribd
Upload
40 views17 pages

Introduction To Neural Network

This document provides an introduction to neural networks and deep learning. It discusses artificial intelligence, machine learning, and deep learning. It provides examples of single and multiple neural networks for supervised learning tasks like housing price prediction. It also covers structured vs unstructured data, different neural network types, and factors enabling deep learning growth like available data, faster computation, and algorithm innovation.

Uploaded by

SANJIDA AKTER
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
40 views17 pages

Introduction To Neural Network

This document provides an introduction to neural networks and deep learning. It discusses artificial intelligence, machine learning, and deep learning. It provides examples of single and multiple neural networks for supervised learning tasks like housing price prediction. It also covers structured vs unstructured data, different neural network types, and factors enabling deep learning growth like available data, faster computation, and algorithm innovation.

Uploaded by

SANJIDA AKTER
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/ 17

Introduction to Neural Network

& Deep Learning

CSE 4237 - Soft Computing


Mir Tafseer Nayeem
Faculty Member, CSE AUST
[email protected]
Introduction
● This is an introductory course that will let you build cutting-edge AI systems.
● Deep learning is also a new "superpower" that will let you build AI systems
that just weren't possible a few years ago.
● This course will teach you how Deep Learning actually works, rather than
using some library without knowing the internal details.
● So after completing it, you will be able to apply deep learning to your own
applications.

2
Artificial Intelligence, Machine Learning, and Deep Learning

3
Example 1 – Single Neural Network
● Given data about the size of houses you want to fit a function that will
predict their price. (Housing Price Prediction)
● It is a linear regression problem because the price as a function of size is a
continuous output.

We know the
prices can never
be negative

Content Credit: Andrew Ng 4


Example 1 – Single Neural Network
“Neuron”
“Input” “Output”
Size (x) Price (y)

● This single neuron implements the function of the housing price prediction.
So we are creating a function called Rectified Linear Unit (ReLU) which
starts at zero.

Content Credit: Andrew Ng 5


Example 2 – Multiple Neural Network
● The price of a house can be affected by other features such as size, number of
bedrooms, zip code and wealth.

Size
Fam
i ly S
ize
#bedrooms
Nearby
Price
{ y
Zip Code
y
u alit
ol Q
ho
Wealth Sc
{

X
Content Credit: Andrew Ng 6
Example 2 – Multiple Neural Network
● By stacking together single neurons, we now have a slightly larger neural network

Size
Fam
i ly S
ize
#bedrooms
Nearby
Price
{ y
Zip Code
y
u alit
ol Q
ho
Wealth Sc
{

X
Content Credit: Andrew Ng 7
Example 2 – Multiple Neural Network
● The role of the neural network is to predicted the price and it will automatically generate
the hidden units.
● We only need to give the inputs x and the output y for a number of examples of your
training set. Neural Network will automatically figure out the hidden layers.

Content Credit: Andrew Ng 8


Example 2 – Multiple Neural Network
● The role of the neural network is to predicted the price and it will automatically generate
the hidden units.
● We only need to give the inputs x and the output y for a number of examples of your
training set. Neural Network will automatically figure out the hidden layers.
● Given enough data (training examples) neural networks are extremely good at figure
out functions that accurately map from x to y.

Input layer and the layer in the


middle (hidden layers) are
densely connected because
every input is connected to
every hidden unit.
Content Credit: Andrew Ng 9
Supervised learning for Neural Network
● In supervised learning, we are given a data set and already know what our
correct output should look like (human annotation / gold label)
● We have some input(x) and we learn a function to map output(y)

10
Content Credit: Andrew Ng
Different types of Neural Networks for Supervised Learning

● CNN or Convolutional Neural Networks (Useful in computer vision)


● RNN or Recurrent Neural Networks (Useful in Speech recognition or NLP)
● Standard NN (Useful for Structured data)
● Hybrid/custom NN or a Collection of NNs types

11
Structured vs Unstructured Data
● Structured data is like the databases and tables.
○ Each of the features (e.g. Size, #bedrooms etc has a well defined meaning)
● Unstructured data is like images, video, audio, and text.
○ Harder for computer to make sense of this kind of data
● Structured data gives more money because companies relies on prediction on its big data.

12
Content Credit: Andrew Ng
Why is deep learning taking off?
Deep learning is taking off due to

● Large amount of data available through the digitization of the society


● Faster computation
● Innovation in the development of neural network algorithm

13
Data
We have a lot of data because the world is using the computer a little bit more

● Mobiles
● IOT (Internet of things)

Scales drives deep


learning progress

14
Content Credit: Andrew Ng
Scales drives deep learning progress
● Data
○ Being able to train a big enough neural network
○ Huge amount of labeled data
● Computation:
○ GPUs.
○ Powerful CPUs.
○ Distributed computing.
● Algorithm:
○ Creative algorithms has appeared that changed the way NN works.
○ For example using RELU function is so much better than using SIGMOID function in training a
NN because it helps with the vanishing gradient problem.

15
The process of training a neural network is iterative.
It could take a good amount of time to train a neural network, which affects your
productivity. Faster computation helps to iterate and improve new algorithm.

16
Content Credit: Andrew Ng
END

17

You might also like