Engineering data has become essential in business-critical systems and applications. Audio, image, real-time video, motion, machine performance metrics, and other sensor-generated data are being combined with traditional business, transactional, and other IT data. This creates enormous opportunities for sophisticated analytics. The flexibility to run those analytics – in massive data sets in IT, as cloud infrastructures, or even as the data are acquired on smart sensors and embedded devices – is enabling organizations in many industries to develop intelligent products, devices, and services that expand the business impact of their data and analytics.
Rich Rovner, VP Marketing, MathWorks spoke to EM more about data analytics driven applications, and adds insights on technical computing.
Could you give us an introduction to MATLAB and analytics support in the latest releases?
We have been investing heavily in data analytics capabilities for the past few years – that is, not just with R2016a but with releases since R2014b. We continue to add new algorithms and improve existing ones for statistics and machine learning, computer vision, neural networks, and other data analytics applications. Here are two recent examples: 1. With 16a, you can train convolutional neural networks, which enables deep learning, using Neural Network Toolbox. 2. In 15a, we released the Classification Learner app in Statistics and Machine Learning Toolbox. It brings classification workflows to the MATLAB community.
In terms of performance, we continue to invest in accelerating MATLAB. We have highly optimized libraries, just-in-time compilation, and a new MATLAB Execution Engine that came out last year. These dramatically speed up MATLAB performance. With parallel computing, MATLAB runs on multiple computing architectures, from multicore desktops to GPUs, computer clusters, and grid and cloud services. There is also a large investment in big data, including MapReduce and Hadoop file structure support introduced in R2014b. So we have a heavy development investment in algorithm performance, big data support, and computing architecture.
What trends do you see in data analytics solutions?
We are seeing MATLAB used for data analytics in a variety of industries like automotive, aerospace, industrial automation, medical devices, retail, life sciences, and healthcare management. Some good examples are predictive maintenance for fleets in the automotive industry, operational logistics and supply chain analytics in industrial automation, and risk modeling in financial services. Small start-ups to major global organizations are able to apply these techniques for data analytics applications.
In the financial services industry, specifically, what are the top data analytics trends?
Clearly, risk management and risk modeling are important business opportunities. Lots of financial organizations around the world use MATLAB data analytics capabilities for this. Specific risk management examples include loan approval analytics and credit scoring analytics. What’s interesting about finance is that they have enormous amounts of data. MATLAB algorithms work across computing architectures. For example, teams prototype their analytic algorithms with a small set of data. Then they deploy the algorithms in a variety of ways into a production IT environment, which is absolutely critical for them. So the flexibility of the computing architecture and the ability to move the algorithm and the computation anywhere are among the technology value drivers that financial services organizations see when working with these types of tools.
What cloud computing trends do you see in 2016?
Probably one of the most visible examples of cloud applications is IoT. These applications get smarter over time with the analytics and machine learning that goes back down to the devices. In many IoT applications there could be thousands of sensors out in the world collecting an enormous amount of data in real time. That data is then aggregated in the cloud and the analytics can be applied against the data aggregator to build a predictive model.
For example, we have a customer creating a smart building management system that collects data on weather patterns, variable energy costs, thermal dynamics, and other factors affecting energy use. It uses analytics and optimization to automatically control the temperature set point in commercial buildings throughout the day. They’re not just changing the set point in the morning and then setting it back in the evening. With this application, they’re helping their customers save 10-25% in energy consumption, which is significant in large-scale enterprises. This is an example where they are using cloud with sensor-based technologies to store the data and perform the analytics. So what the cloud offers is flexibility around where the data needs to reside and, further, where analytics need to reside. The sensors and systems become smart because of the analytics that are created and deployed in those devices. In many cases those are embedded devices and we are seeing more customers incorporate analytics with Model-Based Design for embedded systems workflows.
What’s your forecast for growth of high-performance data analytics solutions?
It’s hard to put a number on it, but we are seeing growth in every industry. We certainly see our products used across automotive, aerospace, industrial automation, financial services, retail, healthcare management, life sciences, etc. The other indicator of this growth is that a lot of people talk about the shortage of data scientists. A data scientist is an individual that has three skills. The first is they are experts in their domain. The second is they understand computation and computing technology and they know how to apply the principals of computation to solve problems. The third is proficiency with statistical and machine learning. It’s rare to find an individual who has all these skills and experience. There are many articles written about a worldwide shortage of data scientists. Now our perspective is to enable those who already have the domain expertise and understand computation. If they are MATLAB users, we bring the data analytics and the domain-specific algorithms to them immediately. In the case of engineering-driven analytics, MATLAB and Simulink are designed around the engineering data analytics workflow. So we basically allow today’s engineers to become today’s data scientists.
What is engineering-driven analytics?
While general analytics is highly pervasive, we see that engineering data is really becoming critical in lots of applications. You can think of traditional business data as coming from transactional systems like financial and retail systems that record the products you might buy at a store or online retailer. That’s one source of data, but another critical source is engineering data generated by smart systems, smart vehicles, drones, unmanned aircrafts, all types of IoT applications where the machine is generating data in an always-on fashion. So massive amounts of engineering data are combined with the traditional business data. And applying analytics to that combination of engineering and business data really produces a better outcome or better predictive model, and a better set of analytics for wherever you want to deploy it.
What is the outlook for MATLAB in the future?
We have a habit not to pre-announce specific features, but generally we are continuing to invest heavily in all these areas. We continue to add algorithms to the toolboxes and MATLAB. We continue to work on performance. We continue to work on big data and computing flexibility. We continue to invest in things like GPU computation. We continue to invest in exploring new domains and enhancing existing ones in every release. One of the nice things in our strategy is twice-yearly releases. It’s easier for customers to absorb rather than waiting a longer period of time and seeing this massive release every one or two years. There are new sets of features, new capabilities customers can learn about, to figure out what’s right for them. Events like MATLAB EXPO offer a great way for customers to stay current and learn about all the new features and capabilities. There’s something good for every MATLAB user.
