Question

Complex deep nets perform best when parallel processing is used. It can be either shared memory or distributed computing. What are the shared memory option available to train a deep net called?

Answer 1

ASICs, FPGAs, and GPUs are few of the shared memory options available to train a deep net.

Explanation :

Yes, Complex deep nets perform best when parallel processing is used. It can be either shared memory or distributed computing.

CPUs with GPUs and numerous cores present within, are the only memory options that can support parallel processing. ASICs, FPGAs, and GPUs contain huge number of cores, and withstand the numerous parallel processings.

• ASICs : Application Specific Integrated Circuits are power efficient, unlike GPUs. One can code them fro the hardware level, so we can say that it is best at customized deep net. Origin2000 module contains three key types of ASIC.

       1. Hub chip

       2. Router chip

       3. Crossbow chip

• FPGAs : Field Programmable Gate Arrays are highly flexible in nature. They are the arrays of numerous discrete digital interconnected sub-circuits at the core. They are customized at logical port level, unlike GPUs static vectorinstruction set. So, we can say that FGPAs are more and fast responsive than GPUs.

• GPUs : Graphical Processing Units have numerous and numerous cores in it. So, they can process huge number of processes in parallel without delay. So, they consume a lot of energy. Their speciality of performing thousands of operations at once, made their way into the AI developing world successfully.

1. For more knowledge on GPUs, check https://brainly.in/question/13333583

    2. Comparision between FPGAs and GPUs at          

        https://brainly.in/question/7490732

Thanks for dropping by. Hope it helps you.