Automated Machine Learning for Predictive Modeling

Auger.AI - The Power Tool For Machine Learning:

Faster, more accurate.

Auger offers the industry's most accurate Automated Machine Learning. It intelligently traverses the infinite space of algorithm/hyperparameter combinations to find the best possible predictive models faster.

Analysts, data scientists and developers can manage the features, target and algorithms attempted for their dataset via Auger’s easy to use model manager. The best algorithm/hyperparameter combinations (trained models) are presented on a leaderboard for evaluation. The user can then deploy their model of choice to a hosted prediction web service.


Two free trial worker hours


Our Impact

If Accuracy Doesn’t Matter, You Don’t Need Machine Learning - Just GuesS!

Auger provides the most accurate Automated Machine Learning solution within bounded timeframes. Auger uses a combination of advanced Bayesian optimization, new warm starting approaches and automatic ensembling to consistently outperform competitors, even on their own cherry-picked datasets. You can replicate these results with publicly available datasets. We even provide tools to help you do that.


In the example above we ran Auger against other leading AutoML tools using a Microsoft’s own 89 datasets chosen to highlight their accuracy. Auger’s error rate on these difficult classification problems was less than 20%, reducing errors from Azure and Google’s AutoML by more than 20%. On more representative datasets the advantage is larger.


See Auger In Action!



How it Works

How the Auger Automated Machine Learning process works.

The user provides structured data via CSV files or relational databases. They identify the target and contributing features and Auger determines whether it's a regression, classification or time series experiment. Auger then preprocesses the data: imputing missing values, removing overly sparse, low variances and over-correlated features, and generating additional features using a variety of heuristics. The features of the dataset are selected and extracted ready to run a prediction experiment.

Auger searches through thousands of algorithms and hyperparameters to find the best performing models and creates ensembles to further improve accuracy. The results are presented in a leaderboard in real-time for the user to view and interpret. The user selects a winning model to deploy. Auger creates a web service prediction endpoint to generate predictions or classifications from new data encountered.


AI Is Eating Software

All companies are now software companies (“Software Ate The World”). All software will be intelligent. This transition needs an enabler. Auger is the tool to make this transition (not a cloud add-on tool).

Auger emphasizes "Automating AutoML". It provides a high level API that allows AutoML to be embedded in larger applications (with a fraction of the code necessary for other AutoML products' APIs). Auger's "Automating AutoML" API (A2ML) is available open source and even supports other AutoML technologies. This makes it easy to validate our claims of higher accuracy, and provides an easier development model than the native APIs of those products.