Understanding Machine Learning and its Potential in the Semiconductor Industry
Andy Steinbach, Founder & CEO, Paradigm Shift AI
After decades of scientific gestation, machine learning has experienced a decade of breakthrough results elevating it to a utility on par with the physical sciences and engineering disciplines. Unlike traditional science, which uses human intelligence to discover theories of underlying mechanisms, modern machine learning sifts through huge amounts of data to automatically discover theories that explain the data and answer key questions, and therefore can be used to make predictions on future data. This has allowed breakthroughs in a range of previously intractable problems, such as machine-based image, speech, and natural language understanding, as well as a huge range of industrial and e-commerce problems related to predictive analytics and optimization. This “new” science - Data Science – is complementary to traditional methods and the recent breakthroughs were largely enabled by developments in so-called Deep Learning (i.e., deep neural networks) by leveraging massive compute power, big data, and a flood of new models, universal algorithms, and the software frameworks to rapidly speed research in this area. This talk presents a technically-motivated overview and survey of these methods and how they work - geared to an audience with physical science backgrounds. We present deep learning applications from various domains, but focus on the potential for using these new techniques in the semiconductor industry for solving a range of challenging problems.
Andy is leading the startup Paradigm Shift AI in its effort to build an Industrial IOT algorithms-platform able to predict the state of industrial machine health, based upon time-domain unstructured sensor data. This capability improves factory production yield and quality control, reduces operating and maintenance expenses, and can improve plant safety. Before this, Andy led the effort at NVIDIA focused on advancing the application of deep learning within the financial services industry, including applications such as fraud detection, credit and insurance underwriting, consumer engagement and recommendations, and investment applications for finding alpha signals in big data. Andy has led product management teams for over a decade, successfully launching numerous category-defining products in the microscopy and semiconductor industries at ZEISS, FEI Company, KLA-Tencor, and Intel. He holds a PhD in device physics form the University of Colorado, Boulder, where his research involved superconducting circuit technology now used in the effort to building quantum computers.