Frameworks, Tools and Libs

A typical Data Scientist/ML Engineer's toolbox.

Basic Python Tools

• Pandas
• Scikit Learn
• Jupyter

Deep Learning / Neural Network

• [Google] Tensorflow
• [Facebook] PyTorch
  • A replacement for NumPy to use the power of GPUs
  • Tensors are similar to NumPy’s ndarrays, with the addition being that Tensors can also be used on a GPU to accelerate computing.
• [Microsoft] CNTK: Cognitive Toolkit
  • https://github.com/Microsoft/cntk https://docs.microsoft.com/cognitive-toolkit/
  • https://www.microsoft.com/en-us/cognitive-toolkit/
  • https://docs.microsoft.com/en-us/cognitive-toolkit/
• Apache MXNet http://gluon.mxnet.io/
• [Deprecated] Theano, Lasagne

Model Deployment

Models may be developed by one language(e.g. Python, R), however your production environment may use another(e.g. Java). One way to bridge the gap is to encode the models in a language/tool-neutral way:

• ONNX: A collaboration between Facebook and Microsoft. Supports Caffe2, PyTorch, and Cognitive Toolkit.
• PMML: XML
• PFA: YAML or JSON

ONNX

Open Neural Network Exchange: AWS, Microsoft, Facebook

ONNX is intended to be a standardized format that will allow deep learning models trained on one framework to be transferred to another framework with minimal extra work.

• https://github.com/onnx/onnx
• http://onnx.ai/