[align=center]
Do you know?
Artificial intelligence and reasoning have very deep roots of the
For example,
The home office large important query system using AI technology has
As the world's most famous master reasoning
"Sherlock HOLMES" (HOLMES) named
It is used to help the police to perform the major cases of crime investigation
And in the study of data scientists daily
Artificial intelligence and reasoning is more closely -
Artificial intelligence workflow consists of four main parts
Extraction, transformation, training and implementation
Carry out the stage in the field of AI in the terms of
Is called a "reasoning"
AI work flow [/align]
AI execution phase (reasoning)
AI process execution phase is the key to the whole work process, at this stage, can directly reflect the many benefits of AI, in the implementation stage, AI model need to be trained and improved, to deploy again for decision (usually deployed to the target edge devices, such as cameras and sensors, etc.),
In many cases, at the same time of reasoning, data scientists also need to constantly assess its accuracy, this trend or feedback analysis is usually not the edges on the device or the Internet of things, but in the use of equipment to capture data and reasoning results are analyzed in the process of complete,
[align=center]
[/align]
Execution phase application scenarios of the range is very wide, almost infinite, covering both real-time analysis and decision-making on use cases, also cover need real-time decision-making and the real-time analysis of use cases, the use of specific case is different, the demand for memory and storage proportion will be different, below, will be through the use of two actual cases,
Real-time analysis
This is a familiar to many of you may use case: a cell phone camera to identify local businesses, or cell phone camera will be used for size measuring equipment (such as the length measurement sofa), the reasoning process is completely on the local device, depending on the use of mobile phone features, in these examples, mobile phone after reasoning doesn't need to keep relevant information, this scenario is real-time, smart phones in the treatment of the camera data, more dependent on memory instead of storage, the use case does not need to even interim storage for a long time, after completion of reasoning wouldn't need data (image),
In a small remote/perform inference on a mobile device, to achieve the perfect balance between high performance and low power consumption of memory is the key, micron technology can provide with various specifications of the low power shape DRAM solution, suitable for mobile devices, the edges of the car and custom equipment,
Real-time analysis after + reasoning analysis
In many business environment, must continue to analyze the process of reasoning, to ensure that the AI reasoning engine in line with expectations, it is a continuous improvement process of the feedback loop, used in such cases, the role of the flash memory is essential, and the faster analysis is performed, the faster the speed of the improved real-time reasoning, in addition, the need of storage capacity and storage location depends on the following factors: keep the data in the short term is on the device, or stored for a long time in equipment outside of the big data in the repository, and finally, in the decision to deploy storage solutions, data retention time is also a factor to consider,
In many use cases in the field of automobile, data may need to be stored on the device to meet regulatory requirements and the self-driving cars at the same time, a large number of real-time reasoning needs to be in a specific time retain all of the data from the various sensors (such as cameras, engine performance data, etc.), so that the transportation security administration to use these data to analyze traffic accident, in fact, by 2020, the typical vehicle is expected to contain more than 300 million lines of code and more than 1 TB (terabytes) storage space!
CodePudding user response:
Learning is great, very good very strong