CS230

Winter 2019
Andrew Ng & Kian Katanforoosh

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Teaching Team

Instructor Instructor Course coordinator Course advisor

Andrew Ng Kian Katanforoosh Swati Dube Younes Bensouda Mourri

Teaching Assistants

Head TAs

Shervine Amidi Daniel Kunin Cristian Aramburu Hoormazd Rezaei Sarah Najmark

Abhijeet Shenoi

Ahmad-reza Momeni Hojat Ghorbani Sagar Honnungar Ishan Patil

Weini Yu
(remote) (remote) (remote)

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Today’s outline

I. What is deep learning? (25min)

II. Course Logistics (15min)

III. Introduction to Deep Learning Applications (20min)

IV. Examples of student projects (10min)

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 What is deep learning?

• The growth in annually published papers in AI has outpaced that of CS.

• A growing number of AI publications by researchers from other scientific
fields (Physics, Chemistry, Astronomy, Material Science, etc.)

[AI Index Report 2018] Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 What is deep learning?

• The number of Scopus papers on Neural Networks had a compound annual

growth rate of 37% from 2014 to 2017.
• It has notably driven the growth of #papers published in ML and CV.

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Today’s outline

I. What is deep learning? (25min)

II. Course Logistics (15min)

III. Introduction to Deep Learning Applications (20min)

IV. Examples of student projects (10min)

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Course Logistics

5 “courses”:
C1: Neural Networks and Deep Learning
C2: Improving Deep Neural Networks
C3: Strategy for Machine Learning Projects
C4: Convolutional Neural Networks
C5: Sequence Models

Example: C2M3: Course 2 Module 3

Schedule is on http://cs230.stanford.edu/syllabus/

We are going to use the Coursera Platform: www.coursera.org

The class forum is on Piazza: piazza.com/stanford/winter2019/cs230

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 One week in the life of a CS230 student

1 module

Complete
Watch videos on Coursera Solve quiz
programming assignments
≈1h ≈20min
≈1-3h

TA sections 15min project

1 week of class ≈ 2 modules + Go to in-class lecture
≈1h20
+ on Fridays + mentorship
≈ 1 hour with TA

Assignments and Quizzes are due every Tuesday at 10am

Do not follow the deadlines displayed on Coursera!!!

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Grading Formula

Grade = 0.02A + 0.08Q + 0.25P a + 0.25M + 0.40P r

A = Attendance
Q = Quizzes
Pa = (Programming) assignments
M = Midterm
Pr = Final-project

Active Piazza participation = 1% bonus

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Late days

Example: For next Tuesday at 10am you have to

complete the following assignments:
- 2 Quizzes:
★ Introduction to deep learning
★ Neural Network Basics

- 2 Programming assignments:
★ Python Basics with Numpy
★ Logistic Regression with a neural network mindset

At 9am on Tuesday: you submit 1 quiz and the 1 PA.

At 3pm on Tuesday: you submit the second quiz.
At 2pm on Wednesday: you submit the second PA.

How many late days did you use? 3 late days

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 Today’s outline

I. What is deep learning? (25min)

II. Course Logistics (15min)

III. Introduction to Deep Learning Applications (20min)

IV. Examples of student projects (10min)

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Projects: SIGN language detection

y=0 y=1 y=2 y=3 y=4 y=5

⎡1 ⎤ ⎡0 ⎤ ⎡0 ⎤ ⎡0 ⎤ ⎡0 ⎤ ⎡0 ⎤
⎢0 ⎥ ⎢1 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢0 ⎥ ⎢0 ⎥ ⎢1 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥
⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢1 ⎥ ⎢0 ⎥ ⎢0 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢0 ⎥ ⎢1 ⎥ ⎢0 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣0 ⎦ ⎣0 ⎦ ⎣0 ⎦ ⎣0 ⎦ ⎣0 ⎦ ⎣1 ⎦

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Assignment: The Happy House

y=0 y=0

can’t enter
the Happy House

y=1 y=1

can enter
the Happy House!

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Assignment: Object detection

[Joseph Redmon, Ali Farhadi: YOLO9000: Better, Faster, Stronger, 2016]

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
[Another fun video generated with YOLOv2 by J. Redmon: https://youtu.be/VOC3huqHrss]
 Projects: others
85% Ramtin

Optimal goalkeeper shoot

Car detection Face recognition
prediction

Music generation Text generation

Art generation

“I love you”

Emojifier Machine translation Trigger word detection

And
Kian Katanforoosh, Andrew Ng, many
Younes more…
Bensouda Mourri
 Assignment: Car detection for autonomous driving

[Deep Learning Specialization] Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 Projects: others
85% Ramtin

Optimal goalkeeper shoot

Car detection Face recognition
prediction

Music generation Text generation

Art generation

“I love you”

Emojifier Machine translation Trigger word detection

And
Kian Katanforoosh, Andrew Ng, many
Younes more…
Bensouda Mourri
[L. Gatys et al.: Image Style Transfer Using Convolutional Neural Networks , 2015] Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 Projects: others
85% Ramtin

Optimal goalkeeper shoot

Car detection Face recognition
prediction

Music generation Text generation

Art generation

“I love you”

Emojifier Machine translation Trigger word detection

And
Kian Katanforoosh, Andrew Ng, many
Younes more…
Bensouda Mourri
 Today’s outline

I. What is deep learning? (25min)

II. Course Logistics (15min)

III. Introduction to Deep Learning Applications (20min)

IV. Examples of student projects (10min)

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Projects: others

Coloring Black&White pictures with Deep Learning

Neural
Network

Predicting price of an object from a picture

Neural
300$
Network

Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri

 Projects: others

Image-to-Image translation with Conditional-GAN

[Hu, Yu & Yu, Spring 2018: http://cs230.stanford.edu/projects_spring_2018/reports/8289557.pdf] Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 Projects: others

LeafNet: A Deep Learning Solution to Tree Species Identification

[Galbally, Rao & Pacalin: Spring 2018, http://cs230.stanford.edu/projects_spring_2018/posters/8285741.pdf] Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 And many more…

Predicting atom energy based on atomic-structure

Visual Question Answering

Cancer/Parkinson/Alzheimer detection

Activity recognition in video

Music genre classification / Music Compression

Accent transfer in a speech

Generating images based on a given legend

Detecting earthquake precursor signals

…
Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri
 To sum up

1. You will learn about wide range of deep learning topics

2. The course is very applied, you will code these applications
3. You have access to mentorship to build an outstanding project in 10 weeks
For next Tuesday (01/15) 10am:
- Create Coursera account and join the private session using the invitation
- Finish C1M1 & C1M2
- 2 Quizzes:
★ Introduction to deep learning
★ Neural Network Basics
- 2 Programming assignments:
★ Python Basics with Numpy
★ Logistic Regression with a neural network mindset

- Find project team-mates and fill-in the Google form that will be posted on Piazza.

This Friday (01/11):

- TA section “Deep Learning Applications”

Download your notebooks after you finished them!

Follow only the website deadlines!
Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri