Automated Machine Learning with AutoGluon Library in Python

Explore tabular data with Autogluon, covering supervised learning tasks like classification and regression, feature versus label, test sets, and evaluation metrics such as accuracy and recall.

Learn supervised learning with historical labeled data, distinguish classification from regression, and use a train-test split to train, evaluate, and tune models like AutoGluon for real-world predictions.

Learn to install autogluon, ingest tabular hotel reservation data, and perform a train-test split to build a classifier predicting cancellations.

Train an AutoGluon tabular predictor on a tabular data set, automate feature engineering for binary classification, fit multiple models, and compare ensemble and single models by accuracy and inference time.

Learn how train test splits and cross validation, including holdout final test sets, enable fair evaluation of models like AutoGluon by testing on unseen data.

Validate models with a holdout test dataset to measure performance on unseen data. Use AutoGluon’s tabular predictor to predict new data, then evaluate accuracy, precision, and recall.

Explore classification metrics in AutoGluon, focusing on accuracy, precision, recall, F1 score, and MCC within a confusion matrix, and learn about the accuracy paradox with imbalanced classes.

Explore model-agnostic feature interpretability in AutoGluon by using permutation shuffling to rank features like lead time and special requests and assess their impact on predictions.

Leverage automatic machine learning with AutoGluon to perform regression on tabular data, predicting Uber ride price from date time and location features with automatic feature engineering and end-to-end training.

Explore regression metrics in Autogluon, comparing mean absolute error, mean squared error, and root mean squared error, and understand how to interpret these in context of continuous predictions.

Learn to constrain Autogluon inference time with infer_limit and optional infer limit batch size, and pass manual hyper parameters via per-model dictionaries for tabular models.

Explore how AutoGluon presets optimize quality, deployment, and interpretability, retrain with refit_full, and clone for deployment to minimize artifacts while evaluating with a leaderboard.

Discover AutoGluon’s automated feature engineering for tabular data, using the auto ml pipeline feature generator and built-in generators, with options to customize a feature pipeline.

Explore how Autogluon unites tabular data, image data, and natural language text in multimodal datasets. Train models that handle mixed inputs and NLP tasks, including sentiment analysis and spam detection.

Explore multimodal natural language classification with AutoGluon on tabular data, using CSV reviews to predict star ratings as a multiclass problem, with automated text feature engineering.

Load a trained AutoGluon predictor, evaluate on the test set, and visualize a confusion matrix for a five-class natural language multiclass task, achieving about 60% accuracy.

Discover how the multimodal predictor combines text, images, and tabular data with transformer models from a model zoo to excel in binary classification on natural language data.

Build a multimodal predictor using image, text, and tabular data with AutoGluon, integrating pet adoption data from Petfinder Kaggle, setting full image paths and train-test splits for prediction.

Explore time series forecasting with Autogluon, building probabilistic forecasts using ETS, ARIMA, and neural networks. See single variate and multivariate time series with known and past covariance.

Learn how to forecast a single variate time series with AutoGluon, focusing on data formatting, train/validation splits, and using known and past covariates for multivariate forecasting.

Format a single variate time series for autogluon by converting the date to a timestamp and defining the item id and target columns.

In part two, train an AutoGluon time series predictor on a sliced dataset to forecast future values using a defined prediction length and the symmetric mean absolute percentage error.

Visualize single variate forecasts in AutoGluon by plotting the mean forecast and the 10%–90% confidence interval with matplotlib, using a reusable plot_predictions function for training and prediction data.

Learn how to forecast with known covariates in AutoGluon time series, preparing past and future covariates such as holidays and day of week, and training a model in a notebook.

learn to use AutoGluon time series predictor with known covariates to forecast future dates, feeding future events like holidays and day of week, and visualize predictions.

Leverage past covariates to improve multivariate time series forecasts with Autogluon. Forecast pollution seven days ahead using a time series data frame and visualize results with confidence intervals.

Explore image classification and object detection with AutoGluon, using the multimodal predictor and Openai's clip, and run experiments in Google Colab to handle image data.

Set up Google Colab notebooks, configure runtime for GPU or TPU, and install Autogluon plus MKV Fool and MMDet for detection to enable multimodal prediction.

Perform image classification with Autogluon multimodal predictor in Google Colab, using a Shopee dataset to train and evaluate on test data, then predict on new images.

Learn how to use OpenAI's clip with AutoGluon for zero-shot image classification by matching images to multiple text prompts, enabling flexible image-text probabilities.