Behind Today’s Trends · Jim Tung MathWorks Fellow 9 July 2014 . Big Data Cloud Computing MOOC...

Behind Today’s Trends The Technologies Driving Change

Jim Tung

MathWorks Fellow

9 July 2014

Big Data

Cloud Computing

MOOC

Industry 4.0

Internet of Things

Wearable Tech

Smart Phones

Software as a Service

Social Computing

Executable Internet

Complex Event Processing

Trends from 2009

Big Data

MOOC

Software as a Service

Social Computing

2014 2009 2010 2011 2012 2013

Google Trends over Time

TRENDS

COME AND GO

R&D Expenditure

(% of GDP)

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

Germany

China

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

Engineers and Scientists in R&D

(per million people)

Long-Term Transformations Driving Changes in Markets

and Technical Applications

Spreading into every industry

Complexifying systems

Speeding up development cycles

Intensifying applied research

Challenging the education systems to produce those skillsets

Our strategy, based on what those transformations imply for the long term

Algorithms in everything

People computing anywhere

Hardware in specialized form factors

Connected chips, devices & systems

ALGORITHMS

IN EVERYTHING

Intel predicts

170 billion transistors

per person

in the world by 2015

1980 1985 1990 1995 2000 2005 2010 2015

200x IN 10 YEARS

Transistors Worldwide

TEXTUAL

&

GRAPHICAL

% Compute Kalman Gain:

W = P*M'*inv(M*P*M'+ R);

% Update estimate

xhat = xhat + W*residual;

% Update Covariance Matrix

P = (eye(4)-W*M)*P*(eye(4)-W*M)' + W*R*W';

Simulink

Stateflow MATLAB

% Predicted state and covariance

x_prd = A * x_est;

p_prd = A * p_est * A' + Q;

% Estimation

S = H * p_prd' * H' + R;

B = H * p_prd';

klm_gain = (S \ B)';

% Estimated state and covariance

x_est = x_prd + klm_gain * (z - H * x_prd);

p_est = p_prd - klm_gain * H * p_prd;

% Compute the estimated measurements

y(:,i) = H * x_est;

Large Collection of Function

Libraries

Mathematical

Modeling

Data Analysis

Algorithm

Development

Computer Vision and

Image Processing

Computational

Finance

Control Systems

Communications

Digital Signal

Processing

Physical Modeling

LTE System Toolbox NEW in Release 2014a

(Golden Reference)

Develop a portion of an

LTE system and verify their

design against the LTE

System Toolbox.

(End-to-end simulation)

See the effect of their

design on overall system

performance.

(Signal generation)

Create an LTE signal to

test a design in the lab and

make performance

measurements.

(Signal recovery) Bring a

recorded LTE signal into

MATLAB and decode it.

Transmitter

Test Waveform

Generation

End-to-end link-level simulation Golden reference

Signal generation and analysis

Signal information recovery

MATLAB System block

Incorporate MATLAB System objects into a Simulink model

Simplify the expression and implementation of algorithms for streaming data

Applications such as audio enhancement, wireless processing, object tracking and radar beamforming

NEW in Release 2013b

Signal processing algorithm

development

Basic math

Advanced digital

signal processing

Rapid iteration for development

and testing

Growing library of algorithms

DSP System Toolbox

Control System Tuner App

Tune fixed-structure

controllers in Simulink

Specify blocks to tune

Add tuning goals

Visualize tuning results

Update the tuned

Simulink blocks from app

NEW in Release 2014a

User-Created Apps

Unified textual & graphical programming

Portfolio of libraries

Apps and resources on MATLAB Central

ALGORITHMS

IN EVERYTHING

HARDWARE

IN SPECIALIZED

FORM FACTORS

RUN YOUR ALGORITHM

Power plants 100s Cars

1,000,000s

Planes

1,000s

HIL Simulator

Real-Time Test System USB Plug-In Device

Microcontroller

CPU-FPGA platforms

Custom ASIC

Microprocessor

Customer Talk

Modellbasierte Entwicklung eingebetteter

Systeme für AUTOSAR mit der

MathWorks-Toolkette

Dr. David Seider, Validas

ExploreToday

• Real-time simulation and

testing

• DSP and vision system

prototyping

• HIL simulation

• Designed for a turnkey

experience Simulink Real-Time Explorer

Build, run, and test real-time applications Simulink Real-Time NEW in Release 2014a

• Real-time simulation and

testing

• DSP and vision system

prototyping

• HIL simulation

• Designed for a turnkey

experience Simulink Real-Time Explorer

Build, run, and test real-time applications Simulink Real-Time NEW in Release 2014a

ExploreToday Track 2

Model-Based Design,

Verifikation, Validierung und Test

Run algorithms on

real-time hardware for

testing and verification

Design iteration and

testing in minutes

Simulink Real-Time Release 2014a

“…with Model-Based Design and rapid real-time prototyping, we

maintain the product development pace that our business demands.”

The Rise of Low-Cost

Hardware for the Masses

Arduino

300,000+ commercially produced

Prices ~$30

Raspberry Pi

2.5+ million shipped

$35

LEGO Mindstorms EV3

$350

TETHERED Write code and communicate

with the board

Low-Cost Hardware Same board … different approaches

EMBEDDED Develop a model and

program the board

Low-Cost Hardware Same board … different approaches

MathWorks.com/hardware

LEGO MINDSTORMS EV3

Samsung Galaxy S4


Hardware support

packages

• Get connected and

running quickly

• 150 packages today,

for Arduino, RaspPi,

iPhone, webcams,

Kinect, and more

MathWorks.com/hardware

LEGO MINDSTORMS EV3

Samsung Galaxy S4


Hardware support

packages

• Get connected and

running quickly

• 150 packages today,

for Arduino, RaspPi,

iPhone, webcams,

Kinect, and more


MATLAB und Simulink in

Forschung und Lehre

Magnus Egerstedt

Khepera 3 MATLAB Simulation

HARDWARE

IN SPECIALIZED

FORM FACTORS Code generation for prototyping and embedded

Run your algorithms directly on real-time hardware

Connecting to low-cost hardware for the masses

CONNECTED

CHIPS, DEVICES,

& SYSTEMS

Internet of Things

1875 1900 1925 1950 1975 2000 2025

“Place” connectivity

PEOPLE 5 billion “People” connectivity via mobile

devices

THINGS 50+ billion

“Thing” connectivity

INFLECTION

POINTS

Growth in Global Connectivity

PLACES 1 billion

© 2010 Ericsson AB – from Joshipura, Arpit, “Infrastructure Innovation - Can the

Challenge be met?” Global Semiconductor Alliance, September 2010

Megabyte 1,000,000 bytes (million)

Gigabyte 1,000,000,000 bytes (billion)

Terabyte 1,000,000,000,000 bytes (trillion)

Petabyte 1,000,000,000,000,000 bytes (million billion)

Exabyte 1,000,000,000,000,000,000 bytes (billion billion)

Big Data in Energy Production

Source: General Electric

Turbine

(machine)

Wind farm

(plant)

Energy Producer

(enterprise)

Analytics Asset

optimization

Operations

optimization

Business optimization

Data

quantity

>100 tags

>6,000 tags

>1M+ tags

Data

frequency

40 milliseconds

1 second

1 second – 10 minutes

17 MB/day

6.3 GB/year

4.1 GB/day

1.5 TB/year

0.7 TB/day

0.25 PB/year

Big Data in Many Industries

ENERGY Smart grid

FINANCE Fraud detection

AUTO Fleet data will

influence vehicle design

AERO Maintenance, reliability

BIOTECH Instrumented humans

Exabyte 1,000,000,000,000,000,000 bytes (billion billion)

Big Data in MATLAB

Memory &

Disk Management

64-bit

Memory Mapped Variables

Disk Variables

Intrinsic Multicore Math

Image Block Processing

Processing Speed

GPU Computing

Parallel Computing

Cloud Computing

Distributed Arrays

Algorithms

Streaming Algorithms

Machine Learning

Big Data in MATLAB

Memory &

Disk Management

64-bit

Memory Mapped Variables

Disk Variables

Intrinsic Multicore Math

Image Block Processing

Processing Speed

GPU Computing

Parallel Computing

Cloud Computing

Distributed Arrays

Algorithms

Streaming Algorithms

Machine Learning

Reduce detail

in the

voiceover and

mention that

it’s also

visualization of

the big data

ExploreToday MasterClass

2^48 - keine Angst vor

großen Datensätzen mit

MATLAB

PROCESSING OPTIONS

• MATLAB RESTful interface to Cluster

• MATLAB Hadoop Streaming

• NoSQL connector (e.g. mongo)

• MATLAB / Java App accessing Cluster

• MATLAB Map-Reduce Components

CONNECTED

CHIPS, DEVICES,

& SYSTEMS Memory management

Computing power

Advanced algorithms

PEOPLE

COMPUTING

ANYWHERE

Cloud as a New Platform

1,000s of applications MILLIONS of users

Terminal - mainframe, mini

HUNDREDS OF MILLIONS

of users 10,000s of applications

PC - LAN, Internet

BILLIONS of users

Cloud – mobile, browser, social, big data

MILLIONS of apps

Source: IDC, 2013

MATLAB Mobile Support for iPhone, iPad & Android

, enhancing your

MATLAB Desktop MATLAB Distributed Computing Server on Amazon EC2

The cloud

The cloud, running

MATLAB, on demand, from anywhere

MATLAB Production Server

Deploy MATLAB analytics into

enterprise IT frameworks

Integrate with databases, webservers

and application servers

Seamless transition from algorithm prototyping

to enterprise-scale analytics without recoding

“This product is a game changer, for sure.”

—Quantlabs

Running in Enterprise IT Environments

MATLAB Production Server

Deploy MATLAB analytics into

enterprise IT frameworks

Integrate with databases, webservers

and application servers

Seamless transition from algorithm prototyping

to enterprise-scale analytics without recoding

“This product is a game changer, for sure.”

—Quantlabs

Running in Enterprise IT Environments


Mathematische Modellierung,

Datenerfassung,

Signalverarbeitung und –analyse

PEOPLE

COMPUTING

ANYWHERE MATLAB on mobile devices

MATLAB on the cloud

MATLAB in enterprise IT frameworks

MOOCs Massive Online Open Courses

Online course

Unlimited participation

Open access via the web

Keynote:

Neues Lernen - MOOC, MATLAB und Simulink

Prof. Dr.-Ing. Georg Fries

Professor der Ingenieurswissenschaften

Hochschule RheinMain

Seize the opportunity to…

• Learn what’s new…

…. and how to apply it

• Engage your colleagues

• Benefit from the experience

ExploreToday

Enjoy the conference!

Behind Today’s Trends · Jim Tung MathWorks Fellow 9 July 2014 . Big Data Cloud Computing MOOC...

Documents

Transcript of Behind Today’s Trends · Jim Tung MathWorks Fellow 9 July 2014 . Big Data Cloud Computing MOOC...