Microsoft unveils Deep Learning Acceleration Platform Brainwave

0
215

Microsoft, a multinational technology behemoth, in Hot Chips 2017 symposium held on August 20–22, unveiled a new deep learning acceleration platform called Project Brainwave for boosting real-time artificial intelligence (AI) platforms. Equipped with an amazingly low latency feature, the new cloud-based programmable system is capable of processing requests in real-time—which essentially implies getting response as soon as you send request.

The cloud-based programmable hardware is especially designed for cloud infrastructures consisting of real-time AI processes. These are becoming increasingly significant in processing vast array of data streams, such as sensor streams, queries, videos, and interactions with users, in real-time.

Batch-free Operations and FPGA Chips Lead to High Throughput

The new machine learning acceleration platform is better any available AI-dedicated hardware, contends the leading software company. The system employs a field-programmable gate array (FPGA) infrastructure that the company has been successfully deploying in its data centers for quite some time. In addition, the architecture of the distributed system allows for batch-free operations, thereby providing real-time interactions in machine learning systems, such as anomaly detection and intruder alert. Since the FPGA chips processes requests as soon as they are streamed by network, this leads to high-throughput.

High-Performance and Scalability in Complex Machine Learning Models

Project Brainwave offers hardware and software that confer flexibility on the design by being functional across a variety of data types. The software supports an assortment of popular deep learning models and provides high-performance in real-time AI operations across a wide spectrum of complex machine learning models. With time, Microsoft is optimistic that the system will offer substantial process improvements in machine learning operations.

The system, currently, isn’t available to the customers, and no timeline has been announced officially so far. However, Microsoft intends to launch the platform for its third-party customers using its Azure cloud platform within the next few months.

Leave a Reply