« Back Automated Machine Learning (Automl) Using Python

By NIlesh Kadam  |  Machine Learning Training  |  On 10/22/2019 11:27:54 PM

Automated machine learning (AutoML) is the process of generating Machine Learning solutions for the data scientist without having to do endless searches on data preparation, data cleansing, model selection,  hyper parameters, ensemble generation parameters and model compression variables. In a typical machine learning application, users have a dataset consisting of input data points to train on. The raw data itself may not be in a form such that all algorithms may be applicable to it out of the box. An expert may have to apply the appropriate data pre-processing, feature extraction, Detection and handling of skewed data and/or missing values and  variable selection, attribute selection or variable subset selection methods that make the dataset more useful for machine leaning model.

Benefits Of Automated Machine learning 

  • To apply Machine learning models users requires different skill sets like domain knowledge ,mathematical and statistical expertise, programming skill Automl helps to reduce these burden.
  • Automl process produces simpler solutions, faster creation of those solutions, and models that often outperform models that were designed by hand.
  • Automl  helps to reduce the time that would take to train ,test, develop a machine learning model.
  • Modern Automl systems also use their experience to improve their performance.
  • Automl can't replace the data scientist domain expertise and task definition but helps him avoid the technical work associated with model development.

AutoML Concepts 

  • Neural Architecture Search 

Neural Architecture search is the process of automating the design of neural networks nodes. usually reinforcement learning are used to design of these networks. In this techniques models are punished for low accuracy and rewarded for high accuracies. This technique are really useful to obtain higher accuracy.

  • Transfer Learning 

Transfer learning is a technique where one uses pre-trained models to transfer what its learned when applying the model to a new but similar dataset. This enables us to obtain high accuracies while using less computation time and power. Neural architecture search is good for problems that require the discovering of new architectures, while transfer learning works best for problems where the datasets are similar to the ones used in pre-training models.

AutoML Solutions In Python 

  • Auto-Keras 

According to the official site:

Auto-Keras is an free source software library for automated machine learning (AutoML). It is designed by DATA Lab at Texas A&M University and community contributors. Auto-Keras provides functions to automatically search for architecture and model hyper parameters of deep learning models. It can be installed using below command

-- pip install auto-keras

  • Auto-Sklearn

Auto-Sklearn is an advanced automated machine learning package derived from Scikit-learn. It’s a replacement for the Scikit-learn estimator. Check below command for installation of this package.

-- pip install auto-sklearn

  • H20

H2O is an open source distributed package in-memory machine learning platform. It is present in both R and Python. This package provides support for statistical & machine learning algorithms.