The 53rd Digital Healthcare Forum
As the applicable area of deep learning algorithm is gradually increasing, the need for faster processing of the vast amount of calculation despite the lack of computing power.
Thus, there are many research activities on accelerating efficiency of deep learning technologies.
The 53rd Digital Healthcare Forum held on June 25 Tuesday, the speaker from Samsung Electronics, Mr. Jin won Lee talked about the research outcomes on the efficient deep learning with the aspect of algorithm and hardware.
If Image Recognition and Image Classification are just the classifications of images, it could be restricted for its application and hard to say it has high efficiency.
High effciency could only be said when it displays the detalied information that we want.
In such sense, while some says the AI will soon substitute the human professionals, but AI will far from catching up the human.
Image recognition requires 5G capacity of GPU memory per one image.
Therefore, as the amount of data to process increases, the importance of securing the memory is simultaneously increases.
In Deep Learning Models, there are issues to solve, such as;
(1) as model size increases, over-the-air updates are getting harder
(2) the issue of speed
(3) the issue of energy efficiency
On this day, the speaker shared his various research activities to solve such problems.
The speaker also shared the You Tube link related to this topic: https://www.youtube.com/channel/UCaJFuqpYqxWuQl9TZky0njA