Upload File

introduction to deep learningcs230.stanford.edu/files/c1m1.pdfandrew ng what you’ll learn courses...

  • Home
  • Documents
  • Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving
27
Introduction to Deep Learning Welcome deeplearning.ai

Upload: trankhue

Post on 28-Apr-2018

219 views

Category:

Documents


3 download

Report

TRANSCRIPT

Page 1: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

IntroductiontoDeepLearning

Welcomedeeplearning.ai

Page 2: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

• AI is the new Electricity• Electricity had once transformed

countless industries: transportation, manufacturing, healthcare, communications, and more

• AI will now bring about an equally big transformation.

Page 3: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

What you’ll learnCourses in this sequence (Specialization):1. Neural Networks and Deep Learning2. Improving Deep Neural Networks: Hyperparameter

tuning, Regularization and Optimization3. Structuring your Machine Learning project4. Convolutional Neural Networks5. Natural Language Processing: Building sequence models

Page 4: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

IntroductiontoDeepLearning

WhatisaNeuralNetwork?deeplearning.ai

Page 5: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Housing Price Prediction

size of house

pric

e

Page 6: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Housing Price Prediction

Page 7: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

#bedrooms

zip code

wealth

size

Housing Price Prediction

!"

!#

!$

!%

y

Page 8: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

IntroductiontoDeepLearning

SupervisedLearningwithNeuralNetworksdeeplearning.ai

Page 9: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Supervised LearningOutput (y) ApplicationInput(x)

Click on ad? (0/1) Online AdvertisingAd, user info

Object (1,…,1000) Photo taggingImage

Text transcript Speech recognitionAudio

Price Real EstateHome features

Chinese Machine translationEnglish

Position of other cars Autonomous drivingImage, Radar info

Page 10: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Neural Network examples

Standard NN Recurrent NN Convolutional NN

Page 11: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Size #bedrooms … Price (1000$s)21041600 2400 �

3000

333�4

400330369 �

540

Structured DataSupervised Learning

User Age Ad Id … Click418018�

27

932429328787312�

71244

101�1

Unstructured Data

Image

Four scores and seven years ago…

Text

Audio

Page 12: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

IntroductiontoNeuralNetworks

WhyisDeepLearningtakingoff?deeplearning.ai

Page 13: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Scale drives deep learning progress

Amount of data

Per

form

ance

Page 14: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

• Data

• Computation

• Algorithms

Idea

Experiment Code

Scale drives deep learning progress

Page 15: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

IntroductiontoNeuralNetworks

AboutthisCoursedeeplearning.ai

Page 16: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Courses in this Specialization

1. NeuralNetworksandDeepLearning2. ImprovingDeepNeuralNetworks:Hyperparameter

tuning,RegularizationandOptimization3. StructuringyourMachineLearningproject4. ConvolutionalNeuralNetworks5. NaturalLanguageProcessing:Buildingsequencemodels

Page 17: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Outline of this Course

Week 1: Introduction

Week 2: Basics of Neural Network programming

Week 3: One hidden layer Neural Networks

Week 4: Deep Neural Networks

Page 18: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

IntroductiontoDeepLearning

SupervisedLearningwithNeuralNetworksdeeplearning.ai

Page 19: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Supervised LearningOutput (y) ApplicationInput(x)

Click on ad? (0/1) Online AdvertisingAd, user info

Object (1,…,1000) Photo taggingImage

Text transcript Speech recognitionAudio

Price Real EstateHome features

Chinese Machine translationEnglish

Position of other cars Autonomous drivingImage, Radar info

Page 20: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Neural Network examples

Standard NN Recurrent NN Convolutional NN

Page 21: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

Size #bedrooms … Price (1000$s)21041600 2400 �

3000

333�4

400330369 �

540

Structured DataSupervised Learning

User Age Ad Id … Click418018�

27

932429328787312�

71244

101�1

Unstructured Data

Image

Four scores and seven years ago…

Text

Audio

Page 22: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

IntroductiontoNeuralNetworks

WhyisDeepLearningtakingoff?deeplearning.ai

Page 23: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Scale drives deep learning progress

Amount of data

Per

form

ance

Page 24: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

• Data

• Computation

• Algorithms

Idea

Experiment Code

Scale drives deep learning progress

Page 25: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

IntroductiontoNeuralNetworks

AboutthisCoursedeeplearning.ai

Page 26: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Courses in this Specialization

1. NeuralNetworksandDeepLearning2. ImprovingDeepNeuralNetworks:Hyperparameter

tuning,RegularizationandOptimization3. StructuringyourMachineLearningproject4. ConvolutionalNeuralNetworks5. NaturalLanguageProcessing:Buildingsequencemodels

Page 27: Introduction to Deep Learningcs230.stanford.edu/files/C1M1.pdfAndrew Ng What you’ll learn Courses in this sequence (Specialization): 1. Neural Networks and Deep Learning 2. Improving

AndrewNg

Outline of this Course

Week 1: Introduction

Week 2: Basics of Neural Network programming

Week 3: One hidden layer Neural Networks

Week 4: Deep Neural Networks


CS230 Deep Learningcs230.stanford.edu/projects_spring_2018/reports/8290634.pdf · alternative of rating food photos' attractiveness to Yelp's published approach that utilized EXIF

Deep Learningcs230.stanford.edu/projects_spring_2019/reports/18681213.pdf · The learning rate we choose is 0.00005 and batch ... Luke Metz, and Soumith Chintala. Unsupervised representation

Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15810323.pdfNetworks and Siamese Neural Networks Marios Andreas Galanis, Vladimir Kozlow ... There is a very large body

CS230 Deep Learningcs230.stanford.edu/projects_spring_2018/posters/8285188.pdfExplore and develop a deep machine learning model that predicts the future price of digital asset such

CS230 Deep Learningcs230.stanford.edu/files_winter_2018/projects/6933119.pdf2 Methods: Quantum mechanics as an optimization problem Carleo [2017] outlines a theoretical formulation

CS230: Lecture 9 Deep Reinforcement Learningcs230.stanford.edu/winter2020/lecture9.pdf · 2020-03-03 · Kian Katanforoosh I. Motivation Human Level Control through Deep Reinforcement

Healthcare DL for - Deep Learningcs230.stanford.edu/files/Deep Learning in Healthcare.pdf · Future of diagnostic access 1. Improve healthcare delivery. CheXNet can help radiologists

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15812470.pdf · upon them by pursuing deep learning techniques. Using techniques like LSTMs, RNNs, and highway networks,

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15813380.pdf · CS230 Final Project: Milestone Topic: Transfer Learning Ajay Sohmshetty (collaboration with Amir

Spring Quarter 2018 Stanford University - Deep LearningCS230: Deep Learning Spring Quarter 2018 Stanford University Midterm Examination 180 minutes Problem Full Points Your Score 1

CS230 Deep Learningcs230.stanford.edu/projects_spring_2019/posters/18673358.pdfThe ability to synthesize subsections of large volumes of texts into a concise, summarative format will

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15811843.pdfincluding flipping, cropping, rotating, and etc. MOM Figure 1. Sample image of Bart Simpson, Homer Simpson

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15813329.pdf · from a 2019 Kaggle Competition*. The latest model achieved 97.2% accuracy against the test set

Simplifying Grocery Checkout with Deep Learningcs230.stanford.edu/projects_fall_2019/reports/26257432.pdf · Simplifying Grocery Checkout with Deep Learning Jing Ning Department of

CS230 Deep Learningcs230.stanford.edu/projects_spring_2018/reports/8316972.pdf · deep learning based methodology to learn a similarity mea- sure between street and shop photos. 2

CS230 Deep Learningcs230.stanford.edu/projects_spring_2019/reports/18678885.pdfis a treasure trove of information, including that related to virality, user sentiment, networks, and

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15812441.pdfStackGAN managed to generate more realistic, higher resolution images by splitting the problem into two

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15810428.pdf · Food image classification has many uses in everyday tasks, ranging from search to tagging. In our

Bankruptcy Prediction by Deep Learningcs230.stanford.edu/projects_winter_2020/reports/32569269.pdf · Deep learning techniques, as the Convolutional Neural Networks (CNN) have only

CS230 Deep Learningcs230.stanford.edu/projects_spring_2019/reports/18681615.pdfStanford University 1050 Arastradero Rd., Stanford, CA kkaganov [ at ] stanford.edu Abstract In order

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15871106.pdf5.0.2 Models Using the retrained Inception ResNet classifier, we ran several experiments to test the

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15812222.pdf · 2019-04-04 · Iterative Cloud Point (ICP) with depth information or iterative model matching architecture

CS230: Lecture 9 Deep Reinforcement Learningcs230.stanford.edu/spring2019/cs230_lecture9.pdf · IV. Deep Q-Learning application: Breakout (Atari) Goal: play breakout, i.e. destroy

CS230 Deep Learningcs230.stanford.edu/projects_winter_2019/reports/15808904.pdf · Erick Cardenas implemented the Convolutional Baseline model, setup and managed the AWS instances

CS230 Deep Learningcs230.stanford.edu/files_winter_2018/projects/6940460.pdf · also begun exploring deep unsupervised learning methods in the healthcare setting. One example includes

CS230 Deep Learningcs230.stanford.edu/projects_fall_2018/reports/12437786.pdf · ANET achieved 0.87 recall rate across all test cases. CS230: Deep Learning, Fall 2018, Stanford University,

CS230: Lecture 9 Deep Reinforcement Learningcs230.stanford.edu/files/Lecture9.pdf · Kian Katanforoosh, Andrew Ng, Younes Bensouda Mourri I. Motivation Human Level Control through

Deep Learningcs230.stanford.edu/files_winter_2018/projects/6912923.pdf · reinforcement learning is used to address resource allocation and management. B. Voice Voice connection IS

Deep Learningcs230.stanford.edu/projects_spring_2019/reports/18681694.pdf · means even if we implemented a perfect predictor for question spans, the maximum improvement is very limited

CS230 Deep Learningcs230.stanford.edu/files_winter_2018/projects/6940506.pdf · OCR focused on historical transcription has been rarely applied on Arabic histor- ical manuscripts

CS230 Deep Learningcs230.stanford.edu/files_winter_2018/projects/6931955.pdfcurrent limitations and potential next steps in section 4. 2 Data We downloaded the entire AffectNet dataset

r g.ai ork - Deep Learningcs230.stanford.edu/files/C1M4.pdf · Andrew Ng Applied deep learning is a very empirical process cost ! # of iterations Idea Experiment Code

CS230: Lecture 9 Deep Reinforcement Learningcs230.stanford.edu/spring2020/lecture9.pdfCS230: Lecture 9 Deep Reinforcement Learning Kian Katanforoosh Kian Katanforoosh I. Motivation

CS230 Deep Learningcs230.stanford.edu/projects_fall_2018/reports/12450133.pdf · using PERCEPTRON, ADALINE, MADALINE and BACK-PROPAGATION models. It turns out to work fairly well

CS230 Deep Learningcs230.stanford.edu/projects_spring_2018/reports/8290329.pdf · 2018-09-28 · Medical diagnostics with retinal images is an active area of research in the deep-