Cognitive computing describes technology platforms that use approaches from artificial intelligence and signal processing. The design is driven by the functioning of the human brain in order to support a decision-making processes in an application field. This is one particular difference to traditional John-von-Neumann design approaches of computing platforms based on the Turing machine principle. Applications of such platforms link data analysis and adaptive page displays to adjust content for particular users. The goal of this design is to be more influential and affective than traditional computing platforms.
Methods and Properties
Selected methods of the platforms are used from the fields of machine learning, natural language processing, speach and vision, human-computer interaction as well as dialog and narrative generation. In contrast to many previous approaches in computing the platform has an interactive property. In other words it interacts easily with users and inherently even interacts with other Cloud services, hardware, or even other users. Adaptivity is a key property of the platforms. This means to deal with information changes and learn new elements when goals and requirements evolve. As the platforms use dynamic big data in real time cognitive computing raise the demand for massive computing power.
A powerful property is to be stateful and iterative that partly implements the analytical capabilities of humans. The platform asks questions if a problem is vaguely defined and finds additional information if needed. In addition cognitive computing platforms recall previous interactions and thus add information that may be suitable for the specific session that is currently active. Another property is to work with contextual information such as time, location, users profile, processes, and goals. This includes the use of data fusion approaches by pooling different sources of information if needed.
Cognitive Computing Applications
In general applications aim to create more accurate models of how the human brain responds to specific stimulus. They aim to develop better models of how the human mind senses and reasons in particular cases. Selected applications are in the field of speech recognition, sentiment analysis, text analysis, graph analytics, face detection. A concrete application example is image and video processing. One example is the extraction of logos on 25 simultanous video streams shown in the video below. Broadcasters would like to know who is advertising at a given time, check have you shown all the ads that you suppose to show based on contractual obligations. The platform is trained on logos to find them in the streams, but it can be trained for other purposes as well.
Another example is to analyse mammograms in order to find and differentiate different kind of tumours. The system would be trained on historical images and one application is then build to find tumours in new previously unseen images. Since this application requires massive computing it is based on powerful High Performance Computing (HPC) servers from Hewlett Packard Enterprises (HPE). In particular this application is using the HPE Apollo HPC servers offerings General Purpose Graphical Processing Units (GPGPUs) together with the HPE CogX toolkit.
More information about cognitive computing
More pieces of Information about cognitive computing can be obtained from the following Video:
Follow us on Facebook: