---
# What is Deep Learning
### Typical ML system
.center[
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# What is Deep Learning
### Typical ML system
.center[
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# What is Deep Learning
### Deep Learning system
.center[
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# Why Deep Learning Now?
- Five decades of research in machine learning
- .grey[CPUs/GPUs/storage developed for other purposes]
- .grey[lots of data from “the internet”]
- .grey[tools and culture of collaborative and reproducible science]
- .grey[resources and efforts from large corporations]
---
# Why Deep Learning Now?
- Five decades of research in machine learning
- CPUs/GPUs/storage developed for other purposes
- .grey[lots of data from “the internet”]
- .grey[tools and culture of collaborative and reproducible science]
- .grey[resources and efforts from large corporations]
.center[
_GPU and TPU_
]
---
# Why Deep Learning Now?
- Five decades of research in machine learning
- CPUs/GPUs/storage developed for other purposes
- lots of data from “the internet”
- .grey[tools and culture of collaborative and reproducible science]
- .grey[resources and efforts from large corporations]
.center[
]
---
# Why Deep Learning Now?
- Five decades of research in machine learning
- CPUs/GPUs/storage developed for other purposes
- lots of data from “the internet”
- tools and culture of collaborative and reproducible science
- resources and efforts from large corporations
.center[
]
---
# DL Today: Vision
###Object detection and segmentation
.center[
]
.credit[Pinheiro et al., arXiv:1603.08695]
---
# DL Today: Vision for control
Quadcopter Navigation in the Forest
Giusti et al, http://rpg.ifi.uzh.ch/docs/RAL16_Giusti.pdf
---
# DL Today: NLP, translation
.center[
Wu et al., arXiv:1609.08144
]
---
# DL Today: NLP, question answering
.center[
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# DL Today: Auto-captioning
.center[
Vinyals et al., arXiv:1411.4555
]
---
#DL Today: Speech-to-Text
.center[
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# DL Today: Style transfer
.center[
Gatys et al., arXiv:1508.06576
]
.credit[from github.com/fzliu/style-transfer]
---
# DL Today: Generative models
.center[
Sampled celebrities [Nvidia 2017]
]
--
.center[
StackGAN v2 [Zhang 2017]
]
.credit[Slide credit: O. Grisel and C. Ollion]
---
# DL Today: Generative models
MusicVAE, Roberts et al. arXiv:1803.05428
---
# DL for RL in games
.center[
]
--
AlphaGo/Zero: Monte Carlo Tree Search, Deep Reinforcement Learning, self-play
.credit[Slide credit: O. Grisel and C. Ollion]
---
# Do It Yourself starts now!
### Python and Jupyter notebook
.center[
https://www.anaconda.com/download/
(Python 3.6 version)
]
--
### Pytorch
.center[
https://pytorch.org/
]
--
### Access to a GPU
.center[
https://colab.research.google.com/
]
---
# And now a first example!