Scribd
Upload
5 views20 pages

Deep Learning Overview and Applications

The document provides an overview of deep learning, explaining its fundamental concepts, including neural networks, loss functions, and the significance of deep representations. It highlights the achievements of deep learning in various fields such as image classification, speech recognition, and autonomous driving. Additionally, it discusses the driving forces behind advancements in deep learning, including hardware improvements, datasets, and algorithmic innovations.

Uploaded by

Ashfaq Ahmed
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
5 views20 pages

Deep Learning Overview and Applications

The document provides an overview of deep learning, explaining its fundamental concepts, including neural networks, loss functions, and the significance of deep representations. It highlights the achievements of deep learning in various fields such as image classification, speech recognition, and autonomous driving. Additionally, it discusses the driving forces behind advancements in deep learning, including hardware improvements, datasets, and algorithmic innovations.

Uploaded by

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

MIRPUR UNIVERSITY OF SCIENCE AND TECHNOLOGY (MUST), MIRPUR

DEPARMENT OF COMPUTER SYSTEMS ENGINEERING(CSE)


Artificial Intelligence
(CSE-471)

Lec [Deep Learning A]

Dr. Ashfaq Ahmed


(Assistant Professor)
What is deep learning?

Artificial Intelligence 3
Learning representations from data

 Input data points

 Examples of the expected output

 A way to measure whether the algorithm is doing


a good job

Artificial Intelligence 4
Sample data

Artificial Intelligence 5
Coordinate change

Artificial Intelligence 6
The “deep” in deep learning

Artificial Intelligence 7
A deep neural network for digit classification

Artificial Intelligence 8
Deep representations learned by a digit-classification model

Artificial Intelligence 9
Understanding how deep learning works

A neural network is parameterized by its weights.

Artificial Intelligence 10
Understanding how deep learning works

A loss function measures the quality of the network’s output.

Artificial Intelligence 11
Understanding how deep learning works

The loss score is used as a feedback signal to adjust the


weights.
Artificial Intelligence 12
What deep learning has achieved so far
 Near-human-level image classification
 Near-human-level speech recognition
 Near-human-level handwriting transcription
 Improved machine translation
 Improved text-to-speech conversion
 Digital assistants such as Google Now and Amazon Alexa
 Near-human-level autonomous driving
 Improved ad targeting, as used by Google, Baidu, and
Bing
 Improved search results on the web
 Ability to answer natural-language questions
 Superhuman Go playing
Artificial Intelligence 13
Early neural networks

• Core ideas of neural networks


• Backpropagation algorithm
• Convolutional neural networks
• dubbed LeNet

Artificial Intelligence 14
Kernel methods

A decision boundary

Artificial Intelligence 15
Decision trees, random forests, and gradient boosting
machines

A decision tree: the parameters that are learned


are the questions about the data. A question
could be, for instance, “Is coefficient 2 in the
data greater than 3.5?”

Artificial Intelligence 16
Back to neural networks

• Geoffrey Hinton at the University of Toronto, Yoshua Bengio at the University of


Montreal, Yann LeCun at New York University, and IDSIA in Switzerland.
• 2011, Dan Ciresan from IDSIA in 2012, a team led by Alex Krizhevsky and advised by
Geoffrey Hinton was able to achieve a top-five accuracy
• European Organization for Nuclear Research, CERN, used decision tree–based methods
for analysis of particle data from the ATLAS detector at the Large Hadron Collider
(LHC); but CERN eventually switched to Keras-based deep neural networks due to their
higher performance and ease of training on large datasets.

Artificial Intelligence 17
Why deep learning? Why now?

Three technical forces are driving advances in machine learning:


 Hardware
 Datasets and benchmarks
 Algorithmic advances

Artificial Intelligence 18
Deep learning properties

 Simplicity

 Scalability

 Versatility and reusability

Artificial Intelligence 19
THANKS

You might also like