Feedback

Release 2017a Highlights

It's March again, which means there at least three things that the world can usually count on: spring begins in the North, autumn begins in the South, and MathWorks puts out another release.

Amongst many other features, Release 2017a provides tools for deep learning and data analytics, and it makes it easy to scale up and run things faster in both MATLAB and Simulink.

Let’s say you have a lot of image data, and you need to design algorithms for object detection. Whether you’re an expert in deep learning, or new to the field, we’ve got you covered. You can train a neural network from scratch, or build off work done by deep learning experts using transfer learning with pre-trained models, which already recognize thousands of objects. You can speed up the training by using GPUs on your multi-core machine, or by scaling up on the cloud. We also provide object detection frameworks so you can use deep learning to locate and classify objects with even better accuracy.

More broadly, machine learning can help you make sense of your data. We incorporated it into tools like the Classification Learner app, which many of you told us made it easy to do classification, so we’ve added the Regression Learner app to give you the same easy and interactive experience. 

Once you’ve prototyped your algorithms and are ready to scale up, you might find that your data doesn’t fit in memory. We introduced tall arrays in 16b, which let you interact with big data using the same familiar syntax that you already know. And now, we’ve expanded tall array support for many machine learning functions. 

Just like in MATLAB, we make it easy to work with lots of data in Simulink. For example, if your simulations use large input signals from MAT files, you can stream the input instead of having to load that data into memory. And now, if you need to run thousands of simulations to explore your design space and test your systems, you can use the new parsim command to easily run those simulations in parallel, again, either on your multicore desktop or all the way up on the cloud.

It’s easier than ever to move to the latest release. You can now upgrade all the models in your Simulink project with one click, and you can integrate C code generated from previous releases with code generated by the current release, which encourages code reuse and reduces verification costs.

17a introduces Automated Driving System Toolbox, which lets you design and test ADAS and autonomous driving systems and includes capabilities like sensor fusion and simulating traffic scenarios. 

With Powertrain Blockset, you can perform fuel economy analyses and accelerate controls development. It contains component libraries for numerous automotive subsystems, as well as fully assembled vehicle models. 

And finally, you can use the new 5G library to develop and simulate candidate algorithms for the emerging 5G wireless standard.

There’s plenty more in this latest release, so download 17a now and check it out.

Product Focus

Recorded: 8 Mar 2017