Scribd
Upload
15 views40 pages

Nividia Slides

Slides de mini curso de Deep Learn da NVIDIA.

Uploaded by

Júnior Martins
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
15 views40 pages

Nividia Slides

Slides de mini curso de Deep Learn da NVIDIA.

Uploaded by

Júnior Martins
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/ 40

FUNDAMENTALS OF

DEEP LEARNING
Part 1: An Introduction to Deep Learning

1
To see lecture notes, make full screen
and click the “notes” button

2
WELCOME!
3
THE GOALS OF THIS COURSE

• Get you up and on your feet quickly


• Build a foundation to tackle a deep learning project
right away
• We won’t cover the whole field, but we’ll get a great
head start
• Foundation from which to read articles, follow tutorials,
take further classes
Part 1: An Introduction to Deep
Learning

Part 2: How a Neural Network


Trains

AGENDA Part 3: Convolutional Neural


Networks

Part 4: Data Augmentation and


Deployment

Part 5: Pre-trained Models

Part 6: Advanced Architectures


AGENDA – PART 1
• History of AI
• The Deep Learning Revolution
• What is Deep Learning
• How Deep Learning is Transforming the World
• Overview of the Course
• First Exercise
HAVE FUN!
HUMAN VS MACHINE LEARNING
Relaxed Alertness

Human Machine
Rest and
Training
Digest

Fight-or-
Prediction
flight
LET’S GET STARTED

9
HISTORY OF AI
10
BEGINNING OF ARTIFICIAL INTELLIGENCE

COMPUTERS ARE MADE EARLY ON, GENERALIZED TURNED OUT TO BE


IN PART TO COMPLETE INTELLIGENCE LOOKED HARDER THAN
HUMAN TASKS POSSIBLE EXPECTED

11
EARLY NEURAL NETWORKS

Inspired by biology

Created in the 1950’s

Outclassed by Von
Neumann Architecture

12
EXPERT SYSTEMS

Highly complex

Programmed by hundreds of engineers

Rigorous programming of many rules


EXPERT SYSTEMS - LIMITATIONS

What are these three images?

14
HOW DO CHILDREN LEARN?

• Expose them to lots of


data
• Give them the “correct
answer”
• They will pick up the
important patterns on their
own

15
THE DEEP LEARNING
REVOLUTION
16
DATA

- Networks need a lot of


information to learn from
- The digital era and the
internet has supplied that
data

17
COMPUTING POWER
Need a way for our artificial “brain” to observe lots of
data within a practical amount of time.

18
THE IMPORTANCE OF THE GPU

A Rendered Image A Neural Network

19
WHAT IS DEEP
LEARNING?
20
DEEP LEARNING FLIPS
TRADITIONAL PROGRAMMING ON
ITS HEAD

21
TRADITIONAL PROGRAMMING
Building a Classifier

1 2 3
Define a set of Program those Feed it
rules for rules into the examples, and
classification computer the program
uses the rules
to classify
MACHINE LEARNING
Building a Classifier

1 2 3
Show model the Model takes Model learns to
examples with guesses, we tell correctly
the answer of it if it’s right or categorize as it’s
how to classify not training. The
system learns
the rules on its
own
THIS IS A FUNDAMENTAL
SHIFT
WHEN TO CHOOSE DEEP LEARNING
Classic
Deep Learning
Programming

If rules are clear If rules are


and nuanced,
straightforward, complex, difficult
often better to to discern, use
just program it deep learning
25
DEEP LEARNING COMPARED TO OTHER AI

Depth and complexity of networks

Up to billions of parameters (and


growing)

Many layers in a model

Important for learning complex rules


HOW DEEP LEARNING IS
TRANSFORMING THE
WORLD
27
COMPUTER VISION

ROBOTICS AND OBJECT SELF DRIVING


MANUFACTURIN DETECTION CARS
G
28
NATURAL LANGUAGE PROCESSING

REAL TIME VOICE VIRTUAL


TRANSLATION RECOGNITION ASSISTANTS

29
RECOMMENDER SYSTEMS

CONTENT TARGETED SHOPPING


CURATION ADVERTISING RECOMMENDATI
ONS
30
REINFORCEMENT LEARNING

ALPHAGO BEATS AI BOTS BEAT STOCK TRADING


WORLD PROFESSIONAL ROBOTS
CHAMPION IN VIDEOGAMERS
GO
31
OVERVIEW OF THE
COURSE
32
HANDS ON EXERCISES

• Get comfortable with the


process of deep learning
• Exposure to different
models and datatypes
• Get a jump-start to tackle
your own projects

33
STRUCTURE OF THE COURSE
“Hello World” of Deep Learning

Train a more complicated model

New architectures and techniques to improve


performance

Pre-trained models

Transfer learning
34
PLATFORM OF THE COURSE

GPU powered cloud server

JupyterLab platform

Jupyter notebooks for interactive coding


SOFTWARE OF THE COURSE

• Major deep learning platforms:


• TensorFlow + Keras (Google)
• Pytorch (Facebook)
• MXNet (Apache)
• We’ll be using TensorFlow and
Keras
• Good idea to gain exposure to
others moving forward

36
FIRST EXERCISE:
CLASSIFY HANDWRITTEN
DIGITS
37
HELLO NEURAL NETWORKS

Train a network to • Historically important and


correctly classify
handwritten digits difficult task for computers

• Get exposed to the example,


Try learning like a
Neural Network and try to figure out the rules
to how it works

38
LET’S GO!

39
40

You might also like