Ai stands for Artificial Intelligence.
HPCSQUARE addresses the space in Machine Learning ranging from
statistical tools, like low data requirement ensemble models(e.g. random forest), support vector machines, naive Bayes, etc, to unsupervised learning techniques like k-means clustering, and also state-of-the-art deep networks.
In contrast to ML algorithms that may require manual features engineering to improve performance, deep networks learn from data (=supervised training) and require less manual work. Some examples are (non exhaustive list): CNNs, YOLO for object detection, Autoencoders, RESNET, Inception networks, GANs, LSTM, Transformers.
We can also address a range of performance options and software engineering options offered by Nvidia (compiler, profiler, NGC,…) and by Microsoft Azure in the area of auto-ml and as-a-service AI.
In context of HPC, ML can be used to significantly reduce the time-to-solution of CFD applications:
- Surrogate Models, e.g. Deep Learning For Steady-State Fluid Flow Prediction In The Advania Data Centers Cloud
- Machine Learning for Fluid Dynamics: great youtube video, showing that you can apply a similar approach to solving for turbulent flow as you do in computer vision; it also shows applications of autoencoders and PCA
- In general, this Sorbonne white paper shows it is all about learning PDEs from data:
- Solving PDE with NNs ‐ Reduced models
- Dealing with partially observed data
- Combining physic models and NNs