Microsoft Open Mind Studio is a development environment for machine learning. It won’t be the first machine learning environment, which gives rise to predictive analysis in dozens of industries. Many industry classifications have years of historical data, including transactions, demographics, and deep enrichment through external datasets, from which models can be built. Training the models results in probabilities that can inform decisions with greater accuracy than previously possible. By varying parameters in algorithms, the models produce greater or lesser probabilities.
Early details of Open Mind Studio features and capabilities liken it to Visual Studio. That might be true in terms of the user interface (UI) layout and multiple functionalities that can be implemented within the confines of the UI and interactive development environment (IDE). Consider the range of developments supported through Visual Studio. The central editor pane takes the majority of the screen space, as it should. In the pane, writing C# code is enhanced with IntelliSense. Libraries are immediately available for review and inclusion. Or, if Visual Studio is hosting a data project, the central pane is a SQL editor with its particular formatting for readability. Servers, databases, tables, and other objects are immediately available in side panes of the IDE. Similarly, when integration services or reports are in development, a graphic design surface supporting a drag-and-connect UI with property panels is the standard UI. Within this context, expect Open Mind Studio to provide an environment for machine learning frameworks and components.
Open Mind Studio is said to feature an IDE supporting data, model management, algorithm development, pipeline processing and scheduling, learning experiments, and life cycle management. The support for these capabilities is platform and product agnostic, and leads to support of Hadoop, Spark, TensorFlow, ChaNa, and other computation frameworks over a federated infrastructure of data storage and resource management. Support for computing platforms ranges from client to cloud, and from device to GPU to distributed direct access memory.
Abstractions lead to integrating deep learning and specialized frameworks, and future code bases. TensorFlow, the open source library from Google, is designed to facilitate research in machine learning, and to make it quick and easy to transition from research prototype to production system. Imagine having a TensorFlow playground implemented within an IDE. For image processing, the Caffe deep learning network can process 60 million images daily through a single GPU. Abstract and integrate the capabilities of SCOPE, the parallel querying capability/language for the cloud storage and computational engine that powers all of Microsoft’s Online Services, including Bing. Combine Caffe with SCOPE through TensorFlow and deep learning in visual assessment can be deployed. The value of such integrations in a single environment is a direction of Open Mind Studio.
It is a standard expectation that access to on-premises and cloud-based data and storage is supported. Adding support for compliance, governance, and quality is needed. Security concerns and alerting will need to be addressed through resource management. Developing learning systems that see, hear, interpret, and interact will build and connect intelligent bots that interact with users on SMS, text, and Cortana voice systems. Parametric batch and iterative processing requires mature scheduling. Life cycle development and deployment suggests that a federated infrastructure might include Team Foundation or Git services. A level of support for CPU and GPU processing control is expected. Direct access to field-programmable gate arrays (FPGA) is needed for Internet of Things (IoT) support of changing data feeds from individual devices.
Microsoft Open Mind Studio will help clients develop data-driven content that will meet and match expectations of targeted customers. Consider a clothing company that advertises heavily via mobile devices and social media networks and brand partners. The need to predict in-store customer traffic and to predict sales on mobile devices is a clear machine learning opportunity. The category of clothes, from sweaters to shirts to pants to accessories, has many attributes. Think of color, sleeve length, collar style, pattern, base color, material, and dozens more. Predicting a proper combination of attributes for a brand in a market is possible when historical data is enriched with external data for demographics and weather and local events.
Microsoft Open Mind Studio is expected to be the next generation of IDE for developers. Having an open mind to technologies and platforms and frameworks is the subtext. #machinelearning is the next popular topic with real value in modern organizations.
About the Author
Dave Leininger has been a Data Consultant for 30 years. In that time, he has discussed data issues with managers and executives in hundreds of corporations and consulting companies in 20 countries. Mr. Leininger has shared his insights on data management, integration, and analytics projects with enterprises in finance, healthcare, and retail for decades. Reach him at Fusion Alliance at dleininger@FusionAlliance.com.