In Release 2020a of its MATLAB and Simulink packages, MathWorks enables engineers to train neural networks in the updated Deep Network Designer app, manage multiple deep learning experiments in a new Experiment Manager app, and choose from more network options to generate deep learning code. R2020a introduces new capabilities specifically for automotive and wireless engineers in addition to hundreds of new and updated features for all users of MATLAB and Simulink.
The new release introduces an enhanced Deep Learning Toolbox that helps users manage multiple deep-learning experiments, keep track of training parameters, and analyze and compare results and code with the new Experiment Manager app. Users can interactively train a network for image classification (see the figure), generate MATLAB code for training, and access pretrained models with the Deep Network Designer app.
Additionally, GPU Coder now offers a more expansive set of networks to implement AI systems on cloud and edge devices, including Darknet-19, Darknet-53, Inception-ResNet-v2, NASNet-Large, and NASNet-Mobile.
Included with the new release are three new products:
• Motor Control Blockset contains a library of motor-control algorithms optimized for generating compact code as well as out-of-the-box support for multiple motor-control hardware kits.
• Simulink Compiler lets engineers generate standalone applications, web apps, and software components from Simulink models, enabling simulations to run without installing Simulink.
• MATLAB Web App Server offers controlled access from a browser to MATLAB web apps deployed across a user’s organization.
There are also major updates, including the ability for automotive engineers to more easily create driving scenarios using imported road data from high-definition geographic maps and optimize shift schedules for performance, fuel economy, and emissions analysis. For wireless engineers, MATLAB support for 5G and WiFi-6 technologies has expanded with additional support for waveform generation and cell detection. And, users can generate C++ classes from MATLAB classes, and the new release introduces message-based communication for service functions and reusable subsystems.
R2020a is available immediately worldwide.