from conventional machine learning to deep learning and beyond.pptx

50
From Conventional Machine Learning to Deep Learning and Beyond by GB Chun Hao, Chang Email: [email protected]

Upload: chun-hao-chang

Post on 13-Jan-2017

99 views

Category:

Data & Analytics


2 download

TRANSCRIPT

Page 1: From Conventional Machine Learning to Deep Learning and Beyond.pptx

From Conventional Machine Learning to Deep Learning

and Beyond

by GB Chun Hao, Chang

Email: [email protected]

Page 2: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Who am I ?I graduated from IDEA lab 4 months ago with a Master degree.

As a junior Machine Learning Engineer, my job is to understand the latest Deep

Learning research on Computer Vision and implement/modify them for the

Company’s needs

The following contents are my extremely limited experiences(less than 4 months) and summary.

Please correct me if you found any error ;)

Page 3: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Overview1. Brief introduction of Deep Learning

2. Two Popular Networks: CNN and RNN

3. Conventional Learning Model vs Deep Learning

4. The Strength of DNN

5. The Flaws of DNN

6. Beyond Regular Machine Learning Tasks

7. Appendix: Hardwares for DNN

Page 4: From Conventional Machine Learning to Deep Learning and Beyond.pptx

What is Deep LearningDeep Learning refers to Deep Neural Network.

It belongs to Neural Networks family, a branch of Machine Learning

Page 5: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Two popular branches in DNN

Convolutional Neural NetworkFor Computer Vision Tasks

Recurrent Neural NetworkFor Sequential Problem

Page 6: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Convolutional Neural NetworkThe most popular Neural Network in Computer VisionCNN is able to capture different levels of feature representation

The following pictures are generated purely by the weights of learned models

Image of the Bird Saxophone

Page 7: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN in Typical Computer Vision Tasks

CNN outperforms hand-crafted methods on Image Classification and Object Detection tasks

Page 8: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Recurrent Neural NetworkRecurrent Neural Network is able to memorize and recall the memory.

RNN are suitable for sequential data:

RNN is good at Time Series and Natural Language Processing. RNN is Turing Complete

Page 11: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Shallow Neural Network

Shallow Neural Network is similar to most of the conventional supervised models

Pros:1.Easy to train and test2.Able to approach any continuous function

Cons:1.Performance depends on well-designed features 2.Difficult to generalize the prediction

Page 12: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Deep Neural NetworkPros:1.Automatically learn the High-level features representation 2.Modularity: DNN can be composed like LEGO bricks3.Able to do Transfer Learning

Cons:1.Requires tons of data for training2.Expensive computation power for training and testing (no CV)

Page 13: From Conventional Machine Learning to Deep Learning and Beyond.pptx

What makes DNN so popular?

It has the three advantages:

1. Self-learned high-level features representations2. Modularity3. Transfer Learning

Why are these so important?

Page 14: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Self-Learned High-Level Feature Representation (⅓)It means that you no longer need hand-crafted "Feature Engineering"

In the past, it takes several experts years to do the Feature Engineering on a specific task

Page 15: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Self-Learned High-Level Feature Representation(⅔)DNN simplify the pipelines of Machine Learning tasks (by removing Feature Engineering), therefore researcher can use similar tools to solve different kinds of tasks.

You can address different kinds of problems with similar DNN models!

Page 16: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Self-Learned High-Level Feature Representation(3/3)

Example: The following LRCN model (CNN + RNN) is able to handle three different kinds of problem: Activity Recognition, Image Description and Video Description

Page 17: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Modularity (¼)

DNN models can be composed just like building LEGO buildings

Page 18: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Modularity (2/4) : For example, we want a DNN model for Object Detection, but we only have a Image Classification DNN model(VGG-16).

We can either construct a complicated pipeline with conventional detectors or compose a bigger DNN model with other DNN modules.

VGG-16

For Image Classification only

Object Detection Task

Modification

Page 19: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Modularity (¾) : From Image Classification to Object detection VGG-16

VGG-16

VGG-16 + ROI Polling Layer

VGG-16 + Region Proposal Network+ ROI Polling Layer

SVM

Selective Search

Selective Search

R-CNN

Fast R-CNN

Faster R-CNN

For Image Classification only

Complicated and slow

End-to-End and Fast

Page 20: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Modularity (4/4) It’s also simple to replace one component in DNN

For example, ResNet-52 outperforms VGG-16 on Image Classification, we can replace the VGG-16 with ResNet-52 in Faster R-CNN in order to improve the overall performance in Object Detection

VGG-16 + Region Proposal Network+ ROI Polling Layer

Faster R-CNN

ResNet-52 + Region Proposal Network+ ROI Polling Layer

Faster R-CNN

VGG-16

ResNet-52Replace

Page 21: From Conventional Machine Learning to Deep Learning and Beyond.pptx

The learned knowledge of one task can be used in another task.

An apple detection model or data can help you to do orange detection!

Transfer Learning(¼)

Page 22: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Transfer Learning(2/4)

Conventional Machine Learning models are difficult to apply Transfer Learning. The model depends on domain specific features and is sensitive to the data distribution. Furthermore, different kinds of model(SVM, decision trees) can not share their weights easily

Difficult

Page 23: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Transfer Learning(¾)Use a pre-trained model and fine-tune it to adapt new domain tasks is a

common practice in DNN.

Example:

VGG-16

ImageNet Model(Classification)

Style Recognition

Pedestrian Detection

Fine-Tuning

Fine-Tuning

Page 24: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Transfer Learning(4/4): Wisdom of the DNN Models

Many researchers not only publish their thesis and source code. The Trained Models are also shared on-line.

You can download the Pre-trained Models, combine them or fine-tune them.

Caffe Model Zoo

Page 25: From Conventional Machine Learning to Deep Learning and Beyond.pptx

DNN Minimize the gaps between areas

Text Data

Voice and other Signals

Image and Videos

Reinforcement Learning

DNN can help you get into other research areas easier!

Page 26: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Are DNN models really that magical?

Hehehe, NO

DNN models are nice tools but not that magical.

Let me give you some clues and my experience

Page 27: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Flaws of DNN (⅓): MS-CNN Compromise on the Data Distribution

Do you find anything conflicting with the advantages of DNN models?

Page 28: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Flaws of DNN (⅔): TSN Compromise on Self-Learning Feature

Do you find anything conflicting with the advantages of DNN models?

Page 29: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Flaws of DNN(3/3): Adversarial Examples

Maybe it’s not a big deal, but Adversarial Examples somehow breaks the belief of “DNN models are similar to human brains”

Page 30: From Conventional Machine Learning to Deep Learning and Beyond.pptx

DNN are still wonderful tools to use

The above problems are not serious and have solutions.

Furthermore, there are tons of researchers are working on better and smarter DNN models.

Page 31: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Beyond Regular Machine Learning TasksExcept for the regular classification and regression tasks. DNN models more than that.

Once you can define and evaluate the task mathematically, you can apply DNN models

Think DNN models as

“Trainable Program that does everything”

Page 33: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN model - Neural Style

Page 34: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN models - Generative Adversarial Network

Page 38: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Combine CNN with RNN

Image + Time Series = Videos and varieties of applications!

Page 39: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN+RNN: LRCN

Page 40: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN+RNN: Image CaptioningDemonstration Attention Mechanism to sequentialize an image by focusing on different spots

RNN module decide which part of image to focus

Page 41: From Conventional Machine Learning to Deep Learning and Beyond.pptx

CNN+RNN: Image Based Question Answering

Page 42: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Deep Q-Networks (Reinforcement Learning)

DNN that implements Q-Learning algorithms

Page 43: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Neural Turing Machine

A Turing machine that is differentiable and can be trained on gradient descent

A DNN model that able to control external storage

Page 46: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Any Question? :)

Page 47: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Hardwares: Nvidia - Biggest Player in DNN

For normal users, the only choice to run DNN is Nvidia video cards

Nvidia will be focusing more on Intelligence Computations and Self Driving Cars in future

Page 48: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Hardwares: Google Tensor Processing Unit

This Google homemade TPU are mysterious and not accessible for outsiders.

Page 49: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Hardwares: Movidius Fathom

A USB stick for DNN computations from Movidus

Movidius was purchased Intel recently

Page 50: From Conventional Machine Learning to Deep Learning and Beyond.pptx

Hardware: Physical Neural Network

Started from 1960, PNN is hardware electrically similar to neural networks.

A BrainChip From NeuromorThings