Python Libraries for Machine Learning

History of Python plotting libraries:

Machine Learning

The table below gives you a rough overview:

Library	description
scikit-learn	general-purpose Machine Learning library
statsmodels	package for training sophisticated linear models and forecasting time series
catboost	efficient implementation of Gradient Boosting that handles many categorical features
scipy	scientific computing, contains hundreds of numerical functions (e.g. numerical optimization and FFT)
surprise	recommender package compatible to scikit-learn
TensorFlow	Deep Learning package developed by Google. Capable of distributed computing
PyTorch	Deep Learning package developed by Facebook. Comparable to TensorFlow
spacy	package for tokenizing and analyzing text
RPy2	adapter package for using R from Python

Helper Libraries

Library	description
NumPy	low-level vectors and matrix data structures, used for linear algebra and handling data in Python
pandas	comfortable data analysis package, built on top of NumPy
imblearn	helper package for learning with imbalanced datasets
Pillow	helper package for basic image manipulation
scikit-image	package with more sophisticated image processing methods
Flask	lightweight web server frequently used to make models available as web services
FastAPI	even more lightweight package to build JSON APIs, built on top of Flask
MLFlow	helps recording and versioning models

Big Data

Library	description
Apache Arrow	an execution layer for in-memory data
Numba	Just-In-Time compilation, faster than NumPy
Paperspace	notebooks, alternative to AWS / Google
SQL	for big data, check AWS Redshift, Google BigQuery, Hive, smartlake
Dask	out-of-memory version of pandas that works for big data sets as well
Spark	distributed computing platform that does many things that pandas and scikit would do
PyTables	flexible tabular data structure for huge datasets

More about scaling up pandas: