Green AutoML

In the realm of Green AutoML, we are dedicated to advancing the sustainability of automated machine learning. Our commitment lies in progressively automating the optimization of machine learning pipelines to enhance algorithm performance efficiently, even for individuals without extensive machine learning expertise. By automating tasks such as preprocessing, algorithm selection, hyperparameter configuration, neural architecture search, and postprocessing, Green AutoML aims to streamline and expedite the model design process for practitioners and researchers alike.

What is Green AutoML?

In contrast to traditional AutoML approaches, Green AutoML places a strong emphasis on sustainability, aiming to reduce the environmental impact associated with automated machine learning processes. At its core, Green AutoML seeks to minimize energy consumption, carbon emissions, and resource usage throughout the AutoML lifecycle.

Key features of Green AutoML include:

  1. Energy-Efficient Algorithms: Green AutoML prioritizes the development and utilization of energy-efficient algorithms and models. By optimizing computational resources and reducing power consumption, it minimizes the carbon footprint associated with AutoML tasks.
  2. Resource Optimization: Green AutoML incorporates strategies to optimize resource usage, such as intelligent scheduling of computational tasks and maximizing hardware utilization. This not only reduces energy consumption but also promotes efficient use of computing resources.
  3. Lifecycle Sustainability: Green AutoML considers sustainability across the entire lifecycle of AutoML, from data preprocessing to model deployment. It encourages the adoption of eco-friendly practices at every stage, including data collection, model training, and inference.
  4. Carbon Footprint Reduction: By deploying energy-efficient algorithms and optimizing resource usage, Green AutoML significantly reduces the carbon footprint of AutoML processes. This aligns with global efforts to mitigate climate change and promote environmental sustainability.
  5. Transparency and Accountability: Green AutoML promotes transparency and accountability in sustainability efforts. It provides mechanisms for tracking and reporting environmental metrics, allowing practitioners to assess the sustainability impact of their AutoML workflows.

Through its focus on sustainability, Green AutoML not only delivers state-of-the-art machine learning solutions but also contributes to a greener and more environmentally responsible future. By embracing energy-efficient algorithms, optimizing resource usage, and promoting transparency, Green AutoML sets a new standard for sustainable AutoML practices.

Practical Relevance

In the domain of Green AutoML, the focus shifts towards sustainability, recognizing the imperative to reduce the environmental impact associated with automated machine learning processes. While automation expedites the development of ML applications, it often comes at a significant cost to energy consumption, carbon emissions, and resource usage. Green AutoML seeks to mitigate these environmental burdens by implementing energy-efficient algorithms, optimizing resource utilization, and promoting eco-friendly practices across the AutoML lifecycle.

Our Contribution


Our library SMAC offers a robust and flexible framework for multi-objective optimization that supports users to figure out a Pareto-optimal machine learning hyperparameter configuration considering more than one objective at the same time, like performance and emissions. With our DeepCAVE package, one can further investigate possible tradeoffs between several objectives and study which hyperparameters are important, e.g., for energy consumption.


Team Lead