automl/auto-sklearn
Auto Sklearn
This is a scikit-learn automated machine learning framework designed to optimize model selection and hyperparameters. It functions as an automated model selector and hyperparameter optimization tool for classification and regression tasks, utilizing an automated ensemble builder to combine high-performing models for increased predictive accuracy.
The system features a distributed search engine that uses Dask for parallel machine learning optimization across CPU cores or clusters. It implements a budget-based evaluation strategy through successive halving to prioritize promising model configurations and manages large-scale searches with resource consumption limits and dataset compression.
The framework covers a broad capability surface including automated data preprocessing for text and sparse data, multi-objective metric optimization, and search space restriction. It also provides monitoring tools for accuracy tracking and model leaderboard visualization to interpret the search process.
The software is available as a pre-configured container environment via Docker to simplify deployment.
Features
- Automated Classification - Provides automated discovery and fitting of the best classification models and hyperparameters within resource constraints.
- Distributed Training - Scales the automated model search and optimization process across multiple CPU cores or clusters using Dask.
- Hyperparameter Optimization Tools - Provides a framework for tuning model settings using advanced techniques like Bayesian optimization and successive halving.
- Automated ML (AutoML) - Optimizes model selection and hyperparameters specifically for scikit-learn pipelines to automate the ML workflow.
- Data Preprocessing - Implements automated cleaning, normalization, and transformation of diverse data types to prepare them for machine learning models.
- Hyperparameter Optimization - Employs automated methods to search for and select the optimal configuration parameters for machine learning models.
- Automated - Automatically searches for and selects the most effective machine learning algorithms and configurations for a task.
- Automated Regression - Automatically finds and fits the best machine learning regression models and hyperparameters.
- Parallel Processing - Distributes automated searches and ensemble building across a cluster using Dask for large-scale parallel processing.
- Distributed Optimization Synchronization - Coordinates parallel model search trials across multiple processes or nodes to accelerate the discovery of optimal configurations.
- Parallel - Utilizes Dask to distribute model optimization trials across multiple CPU cores or clusters.
- Budget-Based Evaluation Strategies - Implements a budget-based evaluation strategy through successive halving to prioritize promising model configurations.
- Ensemble Methods - Automatically builds ensembles by aggregating high-performing models to improve overall predictive performance.
- Machine Learning Training - Allows customization of training through specific metrics, train-validation splits, and resampling strategies.
- Model Ensembling - Combines multiple high-performing models into a single ensemble to maximize predictive accuracy.
- Multi-Target Learning - Implements pipelines capable of predicting multiple continuous target variables simultaneously.
- Preprocessing Pipelines - Allows customization or disabling of the automatic preprocessing pipeline to control data transformations.
- Text Preprocessing Pipelines - Incorporates text-based data processing within the automated machine learning pipeline.
- Sparse Data Structures - Utilizes sparse matrices for input features and target labels to optimize memory for high-dimensional data.
- Multi-Objective Optimization - Balances multiple performance metrics simultaneously to find an optimal trade-off during model selection.
- Algorithmic Resource Limiters - Caps memory and disk space usage during model searches to prevent system crashes and filter inefficient configurations.
- Automated Machine Learning - Automated machine learning toolkit built on scikit-learn.
- AutoML - Automated machine learning toolkit for scikit-learn.
- AutoML Frameworks - Automated machine learning toolkit based on scikit-learn.
- Automated Machine Learning - Automated algorithm and hyperparameter tuning for scikit-learn pipelines.
