I am doing a work that is based on analyzing different Python libraries for Machine Learning.

I chose to analyze Scikit-Learn, Keras, Tensorflow and Pytorch for being the most known ones. The idea was to train different models, both supervised and unsupervised learning, as well as classification and regression. Using different algorithms of each type in each of the libraries and analyzing the difficulties/facilities they present as well as the performance.

My problem arises when trying to perform the same as in sklearn in Tensorflow, since for the latter I find a lot of information for deep learning (neural networks), but not for machine learning algorithms (decision trees, Random Forests, SVM, Linear/Logistic Regression, K-NN, KMeans, Naive-Bayes, etc...).

Is it possible to apply these ML algorithms in tensorflow, or is it only oriented for Deep Learning?



TensorFlow is specifically for implementing neural networks architectures, i.e. Deep Learning.

Imho your objective is too broad, you should focus on some specific type of problem/task for example. It's impossible to compare ML methods in general, because different methods perform differently with different types of problems or different datasets.

Among the frameworks you selected:

  • Scikit-Learn is the only for general statistical ML. There are others but not necessarily in Python.
  • Keras, Tensorflow and Pytorch are all specialized in neural networks.

Anyway in the case of traditional statistical ML, there would be little benefit trying the same algorithm with different frameworks: it's true that there can be differences due to implementation choices, but these are rarely significant.


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.