Are any of the open deep learning toolkits targeted to certain areas, or a all toolkits all purpose toolkits, meaning it is a blackbox for deep learning.

My question comes in regards to Microsoft's CNTK which seem to contain examples of speech and text classification, where others usually just have MNIST or CIFAR...

  • 1
    $\begingroup$ There are examples of both. As of today the best general purpose library is probably pyTorch. $\endgroup$ – Emre May 17 '17 at 17:12

Yes, different toolkits are suited for different purposes given they contain different algorithms. This is due to the lack of a general AI (an AI that is intelligent in all respects). Modern AI has different algorithms which are better suited for different tasks. For example a CNN framework is much better for images than a NN would be, and much better than encoders. However, a LSTM framework would be better suited to capture temporal dependencies in the data.

However, if two frameworks are using the same algorithms they should both be relatively the same. Differences will arise due to the underlining codes being different as can be seen between Keras and TensorFlow. Although, this should be limited.

Worry more about the algorithm you are choosing for your task than the framework. Coding up most machine learning algorithms is not that hard if you want to customize it.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.