The Microsoft Cognitive Toolkit
A framework for building deep learning models using computational graphs
&
| + | Scalability | Scales seamlessly from CPUs to GPUs to clusters, handling massive datasets with ease. |
|---|---|---|
| + | Python/C++ API Support | Includes Python/C++ API support offering flexibility and developer choice |
| + | Tutorials | Onboard quickly with Python notebooks and examples. |
| + | Layers Interface | An easy-to-use interface for building neural networks. |
| + | Custom Layer Support | Allows users to define their custom layers for specific tasks, extending its functionality |
| + | Sequence-to-Sequence with Attention | Supports sequence-to-sequence models with attention mechanisms |
| + | Batch Normalization | Built-in batch normalization for improved training stability. |
| + | Convolutional Neural Networks (CNN) | Powerful for image recognition and classification. |
| + | Feedforward Neural Networks (FFN) | Efficiently processes dense data. |
| + | Recurrent Neural Networks (RNN) | Handles sequential data like speech and text. |
| + | BrainScript Support | Handles multi-dimensional sparse and dense data. |
| + | Automatic Differentiation | Automatically computes gradients, a critical step in training neural networks |
| + | Built-in Data Readers | Built-in data readers optimized for various formats’ data loading and parsing |
System Requirements
Version ↓
| # | Recommended |
|---|---|
| 1 | Ubuntu 16.04 LTS (64 bit) |
| 2 | GNU C++ 5.4.0 |
| 3 | Open MPI v. 1.10.7 |
| 4 | Intel® MKLML library |
| 5 |
|
| 6 | OpenCV v.3.1.0 |
| 7 | zlib v.1.2.8 |
| 8 | libzip v.1.1.2 |
| 9 | OpenJDK 7, 64-bit |
| 10 | Anaconda3 4.1.1 (64 bit) |
| 11 | The presented set of product versions is not restrictive, i.e. CNTK may work well in many other configurations |
| # | Recommended |
|---|---|
| 1 |
|
| 2 | Visual Studio Enterprise 2017 |
| 3 | Microsoft MPI v. 7.0 |
| 4 | Intel® MKLML library |
| 5 |
|
| 6 | OpenCV v.3.1.0 |
| 7 | zlib v.1.2.8 |
| 8 | libzip v.1.1.3 |
| 9 | Java SE Development Kit 8 v1.8.0_131, 64-bit |
| 10 | Anaconda3 4.1.1 (64 bit) |
| 11 | The presented set of product versions is not restrictive, i.e. CNTK may work well in many other configurations |
Alternatives
Artificial Intelligence
TensorFlow Accord.NET AForge.NET Eclipse Deeplearning4j OpenCog
Machine Learning
Massive Online Analysis TensorFlow Apache Mahout Apache Spark Apache MXNet Apache SystemDS Eclipse Deeplearning4j MALLET mlpack OpenCV Orange PyTorch scikit-learn Torch Weka Yooreeka
Deep Learning
TensorFlow Apache MXNet Apache SystemDS Caffe Eclipse Deeplearning4j OpenNN PyTorch Torch Weka
TensorFlow Accord.NET AForge.NET Eclipse Deeplearning4j OpenCog
Machine Learning
Massive Online Analysis TensorFlow Apache Mahout Apache Spark Apache MXNet Apache SystemDS Eclipse Deeplearning4j MALLET mlpack OpenCV Orange PyTorch scikit-learn Torch Weka Yooreeka
Deep Learning
TensorFlow Apache MXNet Apache SystemDS Caffe Eclipse Deeplearning4j OpenNN PyTorch Torch Weka