How is GPU Cloud Critical to AI?

GPU(Graphics Processing Unit) is considered as heart of Deep Learning, a part of Artificial Intelligence. It is a single chip processor used for extensive Graphical and Mathematical computations which frees up CPU cycles for other jobs. [1]

A central processing unit (CPU) is essentially the brain of any computing device, carrying out the instructions of a program by performing control, logical, and input/output (I/O) operations. A graphical processing unit (GPU), on the other hand, has smaller-sized but many more logical cores (arithmetic logic units or ALUs, control units and memory cache) whose basic design is to process a set of simpler and more identical computations in parallel. [2]
What is the quickest way to train a Neural Network?

GPU wins over CPU, powerful desktop GPU beats weak mobile GPU, cloud is for casual users, desktop is for hardcore researchers.

Equipment under test:
CPU 7th gen i7–7500U, 2.7 GHz (from my Ultrabook Samsung NP-900X5N)
GPU NVidia GeForce 940MX, 2GB (also from my Ultrabook Samsung NP-900X5N)
GPU NVidia GeForce 1070, 8GB (ASUS DUAL-GTX1070-O8G) from my desktop
2 x AMD Opteron 6168 1.9 GHz Processor (2×12 cores total) taken from PowerEdge R715 server (yes, I have one installed at home. Not at my home though) [3]

How to train your neural net faster?

Before the boom of Deep learning, Google had a extremely powerful system to do their processing, which they had specially built for training huge nets. This system was monstrous and was of $5 billion total cost, with multiple clusters of CPUs.

Now researchers at Stanford built the same system in terms of computation to train their deep nets using GPU. And guess what; they reduced the costs to just $33K ! This system was built using GPUs, and it gave the same processing power as Google’s system. Pretty impressive right? [4]

GoogleStanford
Number of cores1K CPUs = 16K Crores3GPUs = 18K Crores
Cost$5B$33K
Training timeWeekWeek

 

GPU Vs CPU
GPUCPU
  • Hundreds of simpler cores
  • Very complex cores
  • Thousands of concurrent hardware threads
  • Single-thread performance optimization
  • Maximum floating-point throughput
  • Transistor space dedicated to complex ILP
  • Most die surface for integer and fp units
  • Few die surface for integer and FP unit

 

A method of modifying a three dimensional (3D) volume visualization image of an anatomical structure in real time to separate desired portions thereof. The method includes providing a two dimensional (2D) image slice of a 3D volume visualization image of an anatomical structure, identifying portions of the anatomical structure of interest, and providing a prototype image of desired portions of the anatomical structure. The method then includes using an evolver to evolve parameters of an algorithm that employs a transfer function to map optical properties to intensity values coinciding with the portions of the anatomical structure of interest to generate an image that sufficiently matches the prototype image. If the parameters match the prototype image, the method then includes applying the transfer function to additional 2D image slices of the 3D volume visualization image to generate a modified 3D volume visualization image of the anatomical structure. The method includes using a pattern recognizer to assist the evolver, to classify whether a view is normal or abnormal, and to extract the characteristic of an abnormality if and when detected.

The present invention relates to computer processing of three dimensional medical images and, in particular, to a method and system for modifying a three dimensional (3D) volume visualization image of an anatomical structure in real time to delineate desired portions thereof. [5]

About LifeVoxel.AI

Don't forget to share this post!

Get to know the technology behind LifeVoxel.AI