Artificial Intelligence (AI), a theory where machines perform tasks with intelligence like humans, has been the talk of the town across all industries and for all the right reasons. AI is no longer just used to describe the sophisticated consumer profiling platforms and techniques used by the likes of Google and Facebook, Amazon, etc. There has been widespread investigation on AI applications, research and now live adoption in fields like Healthcare, Manufacturing or Industrial applications, Logistics, Defense & Security, Retail, and many more in order to stay competitive while also taking advantage of the additional business models to generate additional revenue. However, the challenge lies in understanding what AI means in a business context and creating a converged architecture for the specific use case that is economically viable.
At Fujitsu, we observed on our digital co-creation journey with several customers that IoT implementation in industrial enterprise applications is enabling customers to utilize AI to effectively automate their processes. One of the many technologies within the broader AI field, Deep Learning (DL) is seen to be having significant impact in the enterprise AI arena. Manufacturing, Logistics and Healthcare are some of the segments where the availability of sensor data and edge computing has been a key enabler for the adoption of Deep Learning technologies.
There are several definitions of Deep Learning in the market today. Just so that we are on the same page, Deep Learning is a subset of Machine Learning, which deals with multi-layered (or deep) neural networks that learn from vast amounts of data. Practical applications would be deriving the machine’s perception of the data input, which could be text, sounds, images, time-series, etc.
Today, the advances in High Performance Computing, Big Data Technology and the availability of algorithms have shown potential to drive further progress in DL within segments like Automotive/Manufacturing, Logistics and Healthcare. Constructing the right and optimal architecture for the workload is critical and depends on application and use case. For example, if the training data you are dealing with involves detection and classification accuracy, HPC architecture would be beneficial. In addition, if you are dealing with data-intensive application-big data open source software, stacks such as Hadoop, from the likes of Cloudera®, Hortonworks® and MapR®, and self-service analytics capability from Datameer® can be used to construct the appropriate architecture in an enterprise analytics context. Analyst firm Freeform Dynamics has captured the convergence of these technologies and their application in Enterprise AI in a whitepaper. Check it out: “Laying the foundations for Enterprise AI – An architectural approach to deep learning platforms”
At the ISC High Performance 2018, Fujitsu demonstrated an integrated Deep Learning solution that shows the convergence of these technologies to provide improved defect localization and classification, ultimately leading to increased yield, lower risk, and enhanced flexibility in manufacturing product lines. The solution illustrates how businesses could transform their production processes with automated real-time decisions. This integrated offering includes Fujitsu’s image recognition software tools for quality control in production line, AI frameworks, powered by the powerful FUJITSU Server PRIMERGY CX400M4. The portfolio of integrated system PRIMEFLEX from Fujitsu includes PRIMEFLEX for HPC and PRIMEFLEX for Hadoop, which cater to the respective workloads as the names suggest. However, Fujitsu is also releasing blueprints or reference architectures for DL workloads that offer an economically viable, optimal solution for customers who can either utilize their existing ‘HPC or/and Big Data environment’ or jointly co-create a solution from scratch, depending on the specific AI problem identified.
We intend releasing the reference architecture documents based on our experience around AI through digital co-creation projects, providing guidelines to help construct the right integrated solution for DL workload, parameters to be considered, etc. Continue to follow us to learn about the AI reference models or contact us to co-create solutions tailored to your specific business needs around AI.