# R2017b at a Glance

## Release Highlights

### Deep Learning

Use DAG and LSTM networks, label images with an app, perform semantic segmentation, and generate CUDA code for NVIDIA GPUs.

### Data Analytics

Text analytics, customized datastore, more big data plots and algorithms for machine learning, and Microsoft Azure blob storage support.

### Real-Time Software Modeling

Model scheduling effects and implement pluggable components for software environments.

### Verification and Validation

New tools for requirements modeling, test coverage analysis, and compliance checking.

### CUDA Code Generation from MATLAB

Generate CUDA code from MATLAB code for deep learning and embedded vision and run on NVIDIA GPUs.

Moving to the latest release just got easier with the code compatibility report, project-wide upgrades, and cross-release code integration.

## MATLAB Product Family

### MATLAB

• Live Editor support for MATLAB commands with automated, contextual hints for arguments, property values, and alternative syntaxes
• geobubble function for creating interactive maps with bubbles whose size and color vary with data values, and wordcloud function for displaying words at sizes based on frequency or other data
• Code compatibility report to help update code to a newer MATLAB release
• binscatter, plot, and scatter functions for visualizing out-of-memory data in tall arrays
• MATLAB Engine API for running MATLAB code from C++ programs, and passing data between C++ programs and MATLAB using MATLAB Data Array

### Text Analytics Toolbox

• A new product for analyzing and modeling text data

### Optimization Toolbox

• Problem-based optimization using simple expressions to specify objectives and constraints

### Database Toolbox

• Database Explorer app to visually explore relational databases without knowing SQL
• Database Toolbox interface for MongoDB

### Symbolic Math Toolbox

• Unit systems to use physical dimensions, convert between SI and US units, and create custom systems of units

### Model Predictive Control Toolbox

• Approximate QP solution to determine worst-case execution time

## Verification and Validation

• A new product for authoring, managing, and tracing requirements to models, generated code, and test cases

• A new product for verifying compliance with style guidelines and modeling standards

• A new product for measuring test coverage in models and generated code

• Model differencing for comparing and merging differences between models including Stateflow charts and MATLAB Function blocks
• Automatic project creation to convert a folder into a Simulink Project
• Schedulable components for creating compositions of components for modeling scheduling effects in software environments
• Simulation Manager to monitor, inspect and visualize simulation progress and results, including parallel simulation runs

### Stateflow

• Ability to include continuous and periodic Simulink algorithms inside Stateflow states for easier mode switching
•  State transition activity and data value visualization with Sequence Viewer and Simulation Data Inspector
• Transition syntax cues for easier creation of concise, readable diagrams with smart queues and actions, new temporal and message operators, and auto layout

• PID autotuning for implementing embedded PID autotuning algorithm

### Simscape

• Nominal value specification for improving simulation robustness by providing scale of variables to solver
• Operating point management to initialize models from saved operating points

## Signal Processing and Communications

### Antenna Toolbox

• pbstack object for designing and analyzing custom PCB antennas with arbitrary metal-dielectric layers, and generating Gerber files to fabricate printed antennas

### Communications System Toolbox

• Library for simulation of low rate wireless personal area networks technologies using ZigBee® protocol
• Library for simulation of near field communication wireless technologies using NFC protocol

### LTE System Toolbox

• Modeling of physical layer algorithms for 5G channel coding, narrowband Internet of Things (NB-IoT) links, and LTE V2X communication

### LTE HDL Toolbox

• A new product for modeling LTE communications subsystems for FPGAs and ASICs

## Deep Learning

### Neural Network Toolbox

• Directed acyclic graph (DAG) networks for deep learning with more complex architectures to improve accuracy and use popular pretrained models
• Long short-term memory (LSTM) networks for deep learning for time-series classification and prediction
• Deep learning training plots for monitoring training progress for accuracy, loss, validation metrics, and more
• Deep learning image preprocessing for efficiently resizing and augmenting image data for training

### Computer Vision System Toolbox

• Semantic segmentation using deep learning to classify pixel regions in images and evaluate and visualize segmentation results

### GPU Coder

• A new product for generating CUDA code for implementation on NVIDIA GPUs for automated driving and other deep learning applications

## Code Generation

### Embedded Coder

• AUTOSAR support to import AUTOSAR compositions, simulate diagnostic and memory services, and use queued send and receive messages

### Fixed-Point Designer

• Simplified Fixed-Point Tool for easy conversion of Simulink systems to fixed point

### MATLAB Coder

• C code generation for machine learning deployment including discriminant analysis classifier models, linear regression models, support vector machine (SVM) regression models, and ensembles of regression trees

• C++ class generation from Simulink functions including initialize, reset, and terminate events

• ThingSpeak IoT platform interface

### GPU Coder

• A new product for generating CUDA code for NVIDIA GPUs

## Release Notes by Product

* Indicates major update