ai-library-exploration

Goal of the project

To learn, explore and compare various AI Libraries

Comments

Fast AI

• Quite good for deep learning.

Pycaret

• Cannot do deep learning
• Ridiculously simple to use
• Good for tabular data
• Cannot do multi label

MLJar

• Quite similar to pycaret
• Very simple, but only limited to machine learning algorithm

Autosklearn

• limited to sklearn algorithms
• need to do preprocessing yourself

H2O

• Good for tabular data
• I would prefer H2O over pycaret, but both are very easy to use
• Can also do deep learning
• Supports R if that concerns you
• Cannot do multi label

Autogluon

• Autogluon can mix tabular data and text data. This is awesome
• Autogluon has to have GPU for deep learning otherwise it wont run
• Run time is long even with GPU, although i did not explore measures to reduce run time
• Working on text took so look that colab run out of GPU time
• Can do multi label

Conclusion

My prefences are

• Fast AI for deep learning
• H2O for tabular data
• Autogluon is quite flexible, but slow