In an interview with Analytics India Magazine, Prashant Rao, Technical Head at MathWorks India talks about the company’s growing focus on Deep Learning and its expanding user base in India. As Deep Learning capabilities become all pervasive and emerges as a game changer for industries, there is a sign of growing awareness and excitement amongst business leaders. In a freewheeling conversation, Rao shares how MathWorks is making Deep Learning more accessible to a broader audience, the latest release R2018b of MATLAB and Simulink and why Indian engineers and researchers heavily rely on these tools.
Analytics India Magazine: Tell us about your journey at MathWorks and how MathWorks has grown in India over the years?
Sign up for your weekly dose of what's up in emerging technology.
Prashant Rao: I joined MathWorks in 2005 as an Application Engineer focused on MathWorks solutions for FPGA/ASIC Design and Verification. Since then, I moved to India in 2009 to lead the MathWorks India Application Engineering and Training team. We serve engineers, researchers and scientists across the commercial, government and education segments. We have the expertise to engage, support and train customers in India across nearly all application areas including AI, ML, DL, IoT/Predictive Maintenance, Computational Biology, Model-Based Design, Autonomous Driving / ADAS, Robotics and Autonomous Systems, FPGA/ASIC Design and Systems Engineering.
AIM: What is MathWorks vision for making Deep Learning tools more accessible to the developer community in India?
PR: Our goal is to make MATLAB accessible and easy for engineers and scientists to use for deep learning. With tools and functions for managing large data sets, MATLAB also offers specialized toolboxes for working with machine learning, neural networks, computer vision, and automated driving. With just a few lines of code, MATLAB enables researchers and engineers to do deep learning without being an expert, and get started quickly, create and visualize models, and deploy models to servers and embedded devices.
AIM: Can you talk about MathWorks R2018b of MATLAB and Simulink which has enhancements for Deep Learning?
PR: In R2018b, we introduced a new Deep Learning Toolbox, which provides engineers and scientists with a framework for designing and implementing deep neural networks. Now, image processing, computer vision, signal processing, and systems engineers can use MATLAB to more easily design complex network architectures and improve the performance of their deep learning models.
AIM: How does the new Deep Learning toolbox provides the developer community more capabilities to build Deep Learning models effectively and faster?
PR: As Deep Learning becomes more pervasive across industries, there is a need to make it broadly accessible and applicable to engineers and scientists with varying specializations. With Deep Learning Toolbox, novices and experts can learn, apply and conduct advanced research by using the integrated deep learning workflow from research to prototype to production.
We continue to improve user productivity and ease of use for Deep Learning workflows in R2018b through:
- The Deep Network Designer app, which enables users to create complex network architectures or modify complex pretrained networks for transfer learning
- Improved network training performance beyond desktop capabilities by supporting cloud vendors with MATLAB Deep Learning Container on NVIDIA GPU Cloud and the MATLAB reference architectures for Amazon Web Services and Microsoft Azure
- Broadened support for domain-specific workflows, including ground-truth labeling apps for audio, video, and application-specific datastores, making it easier and faster to work with large collections of data
AIM: There is also greater interoperability with MathWorks joining the ONNX community, this will allow the models to be used in other frameworks. Why is interoperability an important feature in tools today?
PR: MathWorks recently joined the ONNX community to demonstrate its commitment to interoperability, enabling collaboration between users of MATLAB and other Deep learning frameworks. Using the new ONNX converter in R2018b, engineers can import and export models from supported frameworks such as PyTorch, MxNet, and TensorFlow.
This interoperability enables models trained in MATLAB to be used in other frameworks. Similarly, models trained in other frameworks can be brought into MATLAB for tasks such as debugging, validation and embedded deployment. In addition, R2018b provides a curated set of reference models that are accessible with a single line of code. Also, additional model importers enable use of models from Caffe and Keras-Tensorflow.
AIM: How does MathWorks Deep Learning toolbox lower the barrier to AI and DL for engineers and developer community in India with easier deep learning workflows?
PR: Deep Learning Toolbox helps to create, analyze, and train deep learning networks. Engineers and scientists can use convolutional neural networks (ConvNets, CNNs) and long short-term memory (LSTM) networks to perform classification and regression on image, time-series, and text data. Apps and plots help you visualize activations, edit network architectures, and monitor training progress.
AIM: Which are some of the top companies around the world use MATLAB for Deep Learning.
PR: From university environments to an industry standard tool for development and analysis, MATLAB is widely used by a range of industries from finance, energy and medical devices to industrial automation, automotive and aerospace in various functions for business-critical applications. For example, MATLAB is used by the finance industry for credit risk modeling and in the automotive sector for automated driving system design. Outside these sectors, there are other interesting use cases, for example UCLA researchers use MATLAB to develop an AI-augmented microscope which can diagnose cancer more accurately. Another use case is in agriculture wherein IntelinAir, a U.S.-based precision agriculture company that focuses on aerial imagery analytics, used manned airplanes to image fields. The data collected was further analysed with MATLAB to enable farmers to prevent an outbreak of weeds and maximise their yield.
AIM: MATLAB Deep Learning Toolbox also faces competition from other popular deep learning frameworks like Tensorflow and Caffe. In that respect, can you share what is the main differentiator of MathWork’s AI products.
PR: Engineers developing Deep Learning inference systems using image, video, text, or signal data need a complete workflow to access datasets and pre-trained models, specify and visualize networks, train with CPUs and GPUs, validate models, and prepare inference models for embedded GPU system development.
- MATLAB contains a full set of capabilities for DL
- MATLAB software is easy for engineers to install and work with
- MATLAB makes it easy to learn and use DL techniques
- MATLAB provides an end-to-end, integrated workflow from research to prototype
AIM: Given how MATLAB and Simulink cut across industries, can you talk about Indian industries which rely heavily on these tools.
PR: We are seeing it already, we are seeing it both in automated driving, we are seeing it in the automotive industry, in the industrial automation and machinery industry and many others as opportunities, awareness and access to tools grows. But we should keep in mind that Deep Learning and AI are becoming pervasive across industries and across the world. Soon, we’ll see it front and centre in every day work because AI doesn’t replace tasks, it enhances them. It has become very good at automating repetitive tasks, maybe not as good at automating much more complicated tasks. I think employees learn the technologies and learn how to apply and figure out which tasks can be automated, and which ones cannot or at least not right now. This will allow engineers and scientists to evolve how they work and move higher and higher up the value chain.
AIM: How has the user segment for Deep Learning Toolbox in India grown over the years?
PR: Researchers, scientists and engineers who are already using MATLAB find it easy to move to deep learning as the capacities in Deep Learning Toolbox is part of their workflows. This is strategically important as engineers need to be able to apply their existing domain expertise. AI and Deep Learning are enhancements to existing domain specific algorithms that could be signal processing, image processing, controls, computer vision etc. What we are doing is delivering AI capabilities that jibe well with the current skill set and workflows in other domains.
AIM: As an aside, what are your predictions for the AI market going into 2019.As an AI leader, can you share which industries are seeing the fastest adoption of AI/ML today in India.
PR: We are going to see AI embedded in things we didn’t even realize had the capacity to accommodate or needed AI. Smartphones will become smarter as will other wireless devices. We’ll also see AI moving into other areas with the growth of sensors whether in clothing, robotics or health care. There is just no end to the types of products that will have intelligence and that we can interact with.