Anaconda Demo: Greener Machine Learning with Intel

With conscious developments towards “Green AI,” a focus has been put on reducing the environmental impact of artificial intelligence systems. With the use of Intel® Extension for Scikit-learn*, an open-source extension package developed by Intel®, we can optimize the performance and energy efficiency of machine learning experiments. With the ever-growing importance of AI in our daily lives, there also comes concerns about potential environmental consequences. AI systems often require substantial energy and computing power, resulting in significant CO2 emissions. To mitigate these issues, the concept of green AI has gained traction, aiming to develop practices that minimize the environmental footprint of AI.

See the video below: Greener Machine Learning with Intel: A Demo

Read the article: Greener Machine Learning Computing with Intel AI Acceleration

Scikit-learn-intelex (Sklearnex), the Intel-accelerated extension for the popular Scikit-learn library, is one such solution for achieving energy-efficient AI. By leveraging specialized software optimizations, sklearnex reduces the compute time required for machine learning experiments. This optimization is crucial because processing large amounts of data demands significant energy resources. The scikit-learn benchmark tool was utilized to evaluate different machine learning algorithms across multiple data analytics frameworks. Various experimental settings were tested, including algorithms such as Principal Component Analysis (PCA) for dimensionality reduction and Support Vector Machines (SVM) for classification and regression, as well as algorithms to test clustering, dimensionality reduction, and more.

The results revealed remarkable improvements in performance and energy efficiency. The Intel-optimized code achieved an 8x faster execution time compared to the non-optimized code. Moreover, the sklearnex settings consumed 3x less CPU energy and resulted in a 2.6x smaller memory footprint on average. These findings demonstrated a direct correlation between energy gains and reduced execution time.

Specific algorithms within the benchmark, such as PCA for dimensionality reduction and SVM for classification and regression, exhibited exceptional energy savings. The efficient optimizations offered by sklearnex enabled these algorithms to process larger datasets more effectively, contributing to substantial reductions in energy consumption.

To put these results into perspective, the estimated CO2 savings from using sklearnex can be significant compared with other common CO2 reduction activities. For example, the use of the non-optimized Scikit-learn package on a single AWS C5.metal instance can generate 0.2 kg of CO2 per hour, whereas sklearnex would generate only 0.04 kg for a similar workload. This reduction of 0.16 kg per hour translates to significant emissions reduction over time.

In conclusion, the importance of adopting green AI practices cannot be overstated and the role of software optimizations like sklearnex in achieving energy-efficient machine learning is critical to the future evolution of AI software systems. While individual savings may seem modest, the cumulative impact across AI development can be substantial. By incorporating green AI principles and leveraging tools like sklearnex, the environmental footprint of AI can be significantly reduced, paving the way for a more sustainable future.

We encourage you to also check out and incorporate Intel’s other AI/ML Framework optimizations and end-to-end portfolio of tools into your AI workflow and learn about the unified, open, standards-based oneAPI programming model that forms the foundation of Intel’s AI Software Portfolio to help you prepare, build, deploy, and scale your AI solutions.