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
