Microsoft Releases New Version of High-Performance, Open-Source, Deep Learning Toolkit

Previously known as CNTK, the Microsoft Cognitive Toolkit version 2.0 allows developers to create, train, and evaluate their own neural networks that can scale…

NVIDIA logo
NVIDIA
•
Brad Nemire
•2 min read•intermediate•
--
•View Original
Deep LearningKeras

Overview

Microsoft has released version 2.0 of the Microsoft Cognitive Toolkit, previously known as CNTK, which enables developers to create, train, and evaluate scalable neural networks across multiple GPUs and machines. The open-source toolkit includes new features like Keras support and model compression tools, enhancing its performance and usability for deep learning applications.

What You'll Learn

1

How to create and train neural networks using the Microsoft Cognitive Toolkit

2

Why using the latest NVIDIA Deep Learning SDK can improve training capabilities

3

When to utilize model compression tools for real-time applications

Key Questions Answered

What are the new features in Microsoft Cognitive Toolkit version 2.0?
The new version includes support for Keras, tools for compressing trained models for real-time use, and performance improvements integrated with the latest NVIDIA Deep Learning SDK. These enhancements allow for more efficient training and evaluation of neural networks across multiple GPUs and machines.
How does the Microsoft Cognitive Toolkit support large datasets?
The Microsoft Cognitive Toolkit is designed to scale across multiple GPUs and machines, making it capable of handling massive datasets efficiently. This scalability is crucial for training complex models that require significant computational resources.
What impact does the Microsoft Cognitive Toolkit have on data processing time?
Using the Microsoft Cognitive Toolkit, the Chesapeake Conservancy was able to create new land cover datasets that are 900 times more informative than existing datasets, significantly reducing the time required for data entry and image processing from months to a fraction of that time.

Key Statistics & Figures

Information density of new datasets
900 times
The new datasets created by the Chesapeake Conservancy using the toolkit have 900 times the information of existing 30-meter resolution datasets.
Area covered by Chesapeake Bay
64,000 square miles
The Chesapeake Bay spans approximately 64,000 square miles across six states and Washington, D.C.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Deep Learning Toolkit
Microsoft Cognitive Toolkit
Used for creating, training, and evaluating neural networks.
Neural Network Library
Keras
Supported by the Microsoft Cognitive Toolkit for building neural networks.
Deep Learning Software
Nvidia Deep Learning SDK
Integrated to accelerate training capabilities within the toolkit.
GPU Architecture
Volta
Optimized for use with the Microsoft Cognitive Toolkit to enhance performance.

Key Actionable Insights

1
Leverage the Microsoft Cognitive Toolkit to enhance your deep learning projects by utilizing its support for Keras and model compression tools.
These features can significantly reduce the time and resources needed for training complex neural networks, making it easier to deploy AI solutions in real-time applications.
2
Consider integrating the latest NVIDIA Deep Learning SDK with the Microsoft Cognitive Toolkit to maximize training performance.
This integration ensures that you are using optimized tools for the most powerful GPU architectures, which can lead to faster training times and better model accuracy.

Common Pitfalls

1
Developers may attempt to write their own neural network algorithms from scratch, which can be inefficient and time-consuming.
As data sizes grow and algorithms become more complex, relying on established tools like the Microsoft Cognitive Toolkit is essential for optimizing performance and reducing development time.