Skip to main content

Purdue-affiliated Startup FWDNXT Designing Low-power Hardware for Deep Learning

[caption id="" align="alignleft" width="343"] The "Snowflake" mobile coprocessor for Deep Learning from Purdue-affiliated startup FWDNXT (via HPCWire).[/caption]

A Purdue affiliated startup called FWDNXT is trying to design low-power mobile deep learning. Eugenio Culurciello, an associate professor at Purdue, says that the "Snowflake" mobile coprocessor "is able to achieve a computational efficiency of more than 91 percent on entire convolutional neural networks..."

I'm not exactly sure what he means by 91% computational efficiency. One of the biggest problems with deep learning right now is the last of agreed-upon benchmarks and means of measuring performance. This could mean 91% of peak theoretical efficiency for the chip in terms of single-precision FLOPs, or 91% of theoretical peak performance on this particular network architecture. Unfortunately, there just isn't any way to know.

The original article is posted below (via HPCWire).

https://www.hpcwire.com/off-the-wire/purdue-affiliated-startup-designing-hardware-software-deep-learning/
Labels:

Popular posts from this blog

Neural Network Dense Layers

Neural network dense layers (or fully connected layers) are the foundation of nearly all neural networks. If you look closely at almost any topology, somewhere there is a dense layer lurking. This post will cover the history behind dense layers, what they are used for, and how to use them by walking through the "Hello, World!" of neural networks: digit classification.
Read more

Arrays of Structures or Structures of Arrays: Performance vs. Readability

It's one of those things that might have an obvious answer if you have ever written scientific software for a vector machine. For everyone else, it's something you probably never even thought about: Should I write my code with arrays of structures (or classes), or structures (or classes) of arrays. Read on to see how both approaches perform, and what kind of readability you can expect from each approach.
Read more

Genetic Algorithms: Mutation

So far in my series on Genetic Algorithms I've covered:  Basic Concepts ,  Encoding ,  Selection , and  Crossover . Those operations will get you most of the way to a functioning optimizer. However, all of the operations we've discussed exert convergence pressure  on the population. That is, selection and crossover have a tendency to draw the population together. The last operator I want to talk about -  mutation - is different. It wants to split the population apart. And as counter intuitive as it seems,  divergence  can be even more important than convergence.
Read more