What s New in MATLAB May 16, 2017

Transcription

1 What s New in MATLAB May 16, The MathWorks, Inc. 1

2 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 2

3 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 3

4 Live Editor Change the way you work in MATLAB Run code, present, share, and collaborate using interactive documents Rich text formatting Interactive plot customization with code generation Interactive equation editing Insert images Hyperlinking See results together with code Side-by-side or inline views Save directly to PDF or HTML 4

5 Redesigned Documentation More examples Easier to navigate Cleaner layout Improved discoverability Added more domain-specific information Release Notes filtering Look for types of changes across a range of releases Highlight only changes that have incompatibility considerations 10

6 Pause or Add Breakpoints During Code Evaluation New feature allows you to troubleshoot problems without specifying breakpoints in advance Pause execution of a program from the Editor Enter debug mode Resume program execution 11

7 New Graphics System New look and feel Easier plot customization Enhanced plot customizations Automatic updating of datetime tick labels Dynamic legends Plot using categoricals New and updated visualizations Animated lines Heatmaps 12

8 Increased Performance Run MATLAB code faster with redesigned execution engine All MATLAB code is now JIT compiled Enhancements Improved script and try/catch performance Execute loops with scalar math faster Construct objects and set properties faster Graphics enhancements Render plots with large numbers of markers faster using less memory 14

9 MATLAB Language Enhancements Programming flexibility Functions in scripts Define local functions in scripts for improved code reuse and readability Property validation for classes Easily control a property's size, class, and value 16

10 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 17

11 Importing and Exporting Data Access online data RESTful, JSON, HTTP, CSV, text, and image data Import improvements Improved performance Improved format detection in readtable and datastore Advanced control over import with detectimportoptions Datetime by default Datastore Text, spreadsheet, images, and custom spreadsheetdatastore for collections of spreadsheet files filedatastore for custom file formats 18

12 Tables Simplifying data management For mixed-type tabular data Holds both data and metadata Supports flexible indexing Built-in functionality (merge, sort, etc.) Database-like functionality Join Stack / Unstack 20

13 Categorical Arrays Storing data with values from finite set of categories For discrete non-numeric data Values drawn from a finite set of possible values ("categories ) More memory efficient than a cell array of strings Can be compared using logical operators Similar to numeric arrays Allows for ordered categories Mathematical compare categories 21

14 Date and Time Arrays Manage time more effectively datetime for representing a point in time duration, calendarduration for representing elapsed time Same data type for computation and display Add, subtract, sort, compare, and plot Customize display formats Nanosecond precision Support for time zones and leap seconds Accounts for daylight saving time 22

15 Timetables Time-stamped tabular data Manage time-stamped tabular data with time-based indexing and synchronization Store column-oriented data variables that have different data types Use dedicated functions to manage timetables Reorganize data Evenly space data by time Align multiple data sets Use any table function with timetables 23

16 Strings The better way to work with text Manipulate, compare, and store text data efficiently >> "image" + (1:3) + ".png" 1 3 string array "image1.png" "image2.png" "image3.png" Simplified text manipulation functions Example: Check if a string is contained within another string Previously: if ~isempty(strfind(textdata,"dog")) Now: if contains(textdata,"dog") Performance improvement Up to 50x faster using contains with string than strfind with cellstr Up to 2x memory savings using string over cellstr 24

17 Data Preprocessing Easier ways to clean up messy data Text manipulation functions More than a dozen functions including contains, endswith, and others *missing functions Find, fill, and remove missing data Options to ignore NaNs with cumulative statistic functions Smooth noisy data with filtering or local regression using smoothdata More easily deal with outliers using isoutlier and filloutliers 25

18 Analyze Data Intuitive data processing Split-Apply-Combine Workflow findgroups splits the data into groups splitapply applies a function to each group, and combines the results Apply functions with tables and timetables using varfun 26

19 Big Data - Tall Arrays Scaling your code to big data Applicable when: Data is columnar with many rows Overall data size is too big to fit into memory Operations are mathematical/statistical in nature Machine Memory Statistical and machine learning applications Hundreds of functions supported in MATLAB and Statistics and Machine Learning Toolbox Tall Data 27

20 Summary of Tall Array Capabilities Provides purpose-built functions for use with data that does not fit in memory Data Access ASCII File Database (SQL) Spreadsheet Custom Files table cell numeric cellstr & string Date & Time categorical Data Munging (100 s of functions) Math Statistics Missing Data Visualization Date/Time String Machine Learning Linear Model Logistic Regression Discriminant Analysis K-means PCA Random Data Sampling Summary Statistics 28

21 Using Tall Arrays Run in parallel on compute clusters MATLAB Distributed Computing Server Local disk Shared folders Databases Compute Clusters Spark + Hadoop Tall arrays MATLAB 100 s of functions supported MATLAB Statistics and Machine Learning Toolbox Run in parallel Parallel Computing Toolbox Run in parallel on Spark clusters MATLAB Distributed Computing Server Deploy MATLAB applications as standalone applications on Spark clusters MATLAB Compiler 29

22 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 30

23 Test Frameworks MATLAB Unit Testing Framework Test your code early and often xunit style framework Script / function / class based testing Works with continuous integration servers Automatic reporting Mocking framework Capture screenshots Performance Testing Framework Time MATLAB code automatically Track performance over time 31

24 Source Control Integration Manage your code from within the MATLAB Desktop Leverage modern source control capabilities GIT and Subversion integration in Current Folder browser Use Comparison Tool to view and merge changes between revisions 32

25 External Interfaces Calling Libraries Written in Another Language From MATLAB Python C/C++ Fortran Java COM components and ActiveX controls RESTful, HTTP, and WSDL web services Calling MATLAB from Another Language Java Python C/C++ Fortran COM Automation server 33

26 App Designer Enhanced design environment Component alignment guides Simpler property inspectors Expanded UI component set Gauges Dials Tabbed interfaces Better code management Object-based code format Better property and method management Code refactoring 34

27 Packaging Apps and User-Created Toolboxes App Packaging Automatically includes all necessary files Documents required products Creates single installation file for easy distribution and installation into the MATLAB apps gallery Toolbox Packaging Create your own toolbox of custom functions and package it as a single installer file Contains all of the code, data, apps, documentation, and examples Included files and folders automatically appear on path when installed View details and uninstall toolboxes with Manage Add-on Toolboxes dialog box Documents required products 37

28 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 38

29 Statistics and Machine Learning Toolbox Making Machine Learning Easy and Accessible Classification Learner App 1-click parallel computing Big data algorithms (using tall arrays) C code generation for predictive models (requires MATLAB Coder) Regression Learner App 45

30 Neural Network Toolbox Deep Learning Construct convolutional neural network (CNN) architecture Extract features from a trained network Perform transfer learning with AlexNet, VGG-16, VGG-19 Access out-of-memory data sets using datastores Train and test using CPUs or multiple GPUs Import forward-looking models from Caffe 46

31 Neural Network Toolbox Deep Learning Training Millions of images from 1000 different categories Prediction Real-time object recognition using a webcam connected to a laptop 47

32 Computer Vision System Toolbox Point cloud processing Structure from motion Estimate the essential matrix and compute camera pose from 3-D to 2-D point correspondences Detect objects using region-based convolution neural networks (R-CNN, Fast R-CNN, Faster R-CNN) 48

33 Computer Vision System Toolbox 3D point cloud processing and visualization 49

34 Automated Driving System Toolbox Algorithm development Sensor Fusion Computer Vision Deep learning Testing and verification Ground Truth Labeling App Traffic scenario generation Sensor Fusion Computer Vision & Deep Learning Visualization tools Ground truth labeling Scenario Generation 50

35 Automated Driving System Toolbox Sensor Fusion and Visualization 51

36 Signal Processing Toolbox Signal Analyzer App with time and frequency domain analysis Analyze signals in time and frequency and navigate through the signals using a panner. Spectral analysis of signals Time domain panning Enhanced command line interface allows specifying input signals and sample rates 52

37 Parallel Computing tall array support for big data Measure data sent to workers using ticbytes and tocbytes Cloud offerings with K80-equipped GPUs Simplified parallel Simulink simulations using parsim Send data to client using DataQueue and PollableDataQueue Train a single deep learning network with multiple CPUs or multiple GPUs 53

38 Other Toolbox Enhancement Highlights Optimization Toolbox Mixed Integer Linear Programming Mapping Toolbox Web map display with dynamic base maps Database Toolbox Analyze large data using a DatabaseDatastore and tall table capabilities Retrieve graph data from Neo4j Graph Database Econometrics Toolbox Bayesian Linear Regression Vector Autoregressive Models (VAR) with Exogenous Inputs (VARX) MATLAB Report Generator Document Object Model (DOM) API for creating customized reports 54

39 New Toolboxes Automated Driving System Toolbox Robotics System Toolbox Audio System Toolbox Vision HDL Toolbox Simulink Test Risk Management Toolbox Simscape Fluids Antenna Toolbox Phased Array System Toolbox WLAN System Toolbox LTE System Toolbox 5G library for LTE available 55

40 Agenda MATLAB Foundation Working with Data Building & Sharing MATLAB Applications Application Specific Enhancements Summary and Wrap-up 56

41 Other Topics Data Analytics/Machine Learning Parallel Computing Image Processing and Computer Vision Deep Learning Controls Analysis State Machines Verification & Validation Test Automation Interfacing with ROS 57