Deep Learning

class: center, middle, inverse, title-slide

# <img src="figures/Thanksgiving-Day.jpg" /> Deep Learning
### <font size="5"> Son Nguyen </font>

---

<style>

.remark-slide-content {
  background-color: #FFFFFF;
  border-top: 80px solid #F9C389;
  font-size: 20px;
  font-weight: 300;
  line-height: 1.5;
  padding: 1em 2em 1em 2em
}

.inverse {
  background-color: #696767;
  border-top: 80px solid #696767;
  text-shadow: none;
  background-image: url(https://github.com/goodekat/presentations/blob/master/2019-isugg-gganimate-spooky/figures/spider.png?raw=true);
	background-position: 50% 75%;
  background-size: 150px;
}

.your-turn{
  background-color: #8C7E95;
  border-top: 80px solid #F9C389;
  text-shadow: none;
  background-image: url(https://github.com/goodekat/presentations/blob/master/2019-isugg-gganimate-spooky/figures/spider.png?raw=true);
	background-position: 95% 90%;
  background-size: 75px;
}

.title-slide {
  background-color: #F9C389;
  border-top: 80px solid #F9C389;
  background-image: none;
}

.title-slide > h1  {
  color: #111111;
  font-size: 40px;
  text-shadow: none;
  font-weight: 400;
  text-align: left;
  margin-left: 15px;
  padding-top: 80px;
}
.title-slide > h2  {
  margin-top: -25px;
  padding-bottom: -20px;
  color: #111111;
  text-shadow: none;
  font-weight: 300;
  font-size: 35px;
  text-align: left;
  margin-left: 15px;
}
.title-slide > h3  {
  color: #111111;
  text-shadow: none;
  font-weight: 300;
  font-size: 25px;
  text-align: left;
  margin-left: 15px;
  margin-bottom: -30px;
}

</style>

# Agenda

- Deep learning revolution

- Python Example

- Apple diseases Detection: CNN vs. Gradient Boosting 
  
  - Pneumonia X-ray detection
  
  - GAN

- Define neural networks

---
# Deep Learning

![](images/cnn6.png)

---
# Example

https://www.youtube.com/watch?v=l82PxsKHxYc&t=47s

---

# 2009: ImageNet Competition

.pull-left[

- Launched in 2009

- 15 million high-resolution images

- 22,000 categories]

.pull-right[

![](images/cnn1.png)
]
---
# AlexNet

.pull-left[

- Dramatic performance gain

- No Teams get better than 25% accuracy by 2012

- AlexNet was imtroduce in 2012 with 15.3% accuracy

- No team did worst than 25% since

]

.pull-right[
![](images/cnn2.png)
]

---
# AlexNet

![](images/cnn8.png)

---
# CNN

![](images/cnn3.png)

---
# 2014: DeepFace - Facebook

- In 2014, Facebook - DeepFace can identify faces with 97.35% accuracy, beating rivals human performance

![](images/deep_face.png)

---
# 2016: AlphaGo - Google

- In 2016, AlphaGo became the first computer program to defeat a professional human Go player

- the first to defeat a Go world champion,

- arguably the strongest Go player in history.

- AlphaGo was one of the Breakthrough of the Year by Science.

---
# 2016: Google Translate

- In 2016, Google Translate Deep Learning blew other language translation techniques out of the water

![](images/google_translate2.png)

---
# Google Assistant

https://ai.googleblog.com/2018/05/duplex-ai-system-for-natural-conversation.html

![](images/google_assistant.png)

---
# 2017: Face ID - Apple

- In 2017, Apple switched to use Deep Learning for face recognition

> We had to completely rethink our approach so that we could take advantage of this paradigm shift. https://machinelearning.apple.com/research/face-detection

---
# Apple Siri

https://machinelearning.apple.com/research/siri-voices

![](images/siri.jpg)

---
# Apple - Deep Learning

![](images/cnn5.png)

---
# Amazon

![](images/amazon.jpg)

---
class: inverse, middle, center
# Academia

---
# Academia

---
# 2018: Alan Turing Award

---
class: inverse, middle, center

# Big Tech Companies Interested

---
# Google

- In 2014, Google acquired DeepMind.

- https://deepmind.com/research/case-studies/alphago-the-story-so-far

- https://en.wikipedia.org/wiki/DeepMind

- https://www.tensorflow.org/about

---
# Tensorflow

- Google developed TensorFlow in 2015

- particularly focuses on training and inference of deep neural networks.

- Company uses Tensorflow

![](images/cnn_tensor_flow.png)

- https://www.tensorflow.org/about

---
# Google Brain

- Google Brain is a deep learning artificial intelligence research team at Google.

- Formed in the early 2010s

- https://research.google/teams/brain/

---
# Apple

- Deep Learning Team:

https://www.apple.com/jobs/us/teams/machine-learning-and-ai.html

---
# Facebook: Pytoch

- In 2016, Facebook developed PyTorch, an open source deep learning framework

---
# Microsoft

Deep Learning Group: https://www.microsoft.com/en-us/research/group/deep-learning-group/

---
class: inverse, center, middle

# But...

---
# But...

- Neural Network was invented a long time ago

- Why now?

---
# CNN

---
# Big Data

---
# Big Data

---
# Big Data

---
# Powerful Computing

- https://en.wikipedia.org/wiki/History_of_supercomputing

---
# Powerful Computing
<img src="images/computing.png" style="display: block; margin: auto;" />

---
# Powerful Computing

- If AlphaGoZero was trained in 1 day in 2020, how long would it take to train AlphaGoZero in 2012?

- More than 821 years!

---
# Cheaper Computing

---
# CNN

![](images/cnn4.png)

---
class: middle, inverse, center

# Neural Network

---
# A linear Model is a Neural Network

---
# Number of Parameters

![](images/cnn_pam.png)

- FLOP: Number of Operations
---
# Deeper and Deeper

![](images/cnn_alexnet.png)