H2O4GPU - Collection of GPU Algorithms

H2O4GPU is an open source, GPU-accelerated machine learning package with APIs in Python and R that allows people to make use of GPUs to build advanced machine learning models. H2O provides a variety of popular algorithms including Gradient Boosting Machines (GBM’s), Generalized Linear Models (GLM’s), and K-Means Clustering. According to H2O.ai, these algorithms offer up to a 40X speedup when compared to CPUs.

Here are specific benchmarks from a recent H2O4GPU test by H2O.ai:

• • • • More than 5X faster on GPUs as compared to CPUs
      • Nearly 10X faster on GPUs
      • More than 40X faster on GPUs

Generalized linear models are regression models built on exponential families that have found wide practical application.

H2O’s Gradient Linear Model:

• • • 1. Framework utilizes Proximal Graph Solver (POGS)
      2. Solvers include Lasso, Ridge Regression, Logistic Regression, and Elastic Net Regularization
      3. Improvements to original implementation of POGS:
         • • • Full alpha search
             • Cross Validation
             • Early Stopping
             • Added scikit-learn-like API
             • Supports multiple GPU’s

Gradient Boosting Machine (for Regression and Classification) is a forward learning ensemble method. The guiding heuristic is that good predictive results can be obtained through increasingly refined approximations. H2O’s GBM sequentially builds regression trees on all the features of the dataset in a fully distributed way – each tree is built in parallel.

• • • 1. Based on XGBoost
      2. Raw floating point data — binned into quantiles
      3. Quantiles are stored as compressed instead of floats
      4. Compressed quantiles are efficiently transferred to GPU
      5. Sparsity is handled directly with high GPU efficiency
      6. Multi-GPU enabled by sharing rows using NVIDIA NCCL AllReduce

K-means clustering is a type of unsupervised learning, which is used when you have unlabeled data. The goal of this algorithm is to find groups in the data, with the number of groups represented by the variable K. The algorithm works iteratively to assign each data point to one of K groups based on the features that are provided. Data points are clustered based on feature similarity.

H2O’s K-means Clustering Algorithms

1. 1. 1. 1. Based on NVIDIA prototype of k-Means algorithm in CUDA
         2. Improvements to original implementation:
            1. 1. • Significantly faster than scikit-learn implementation (50x) and other GPU implementations (5-10x)
                  • Supports multiple GPUs

H2O’s GPU-accelerated machine learning algorithms, along with NVIDIA, have allowed enterprises to apply the data science techniques. These solutions can help to solve your enterprise problems with speed, and at scale. Customers everywhere are using massively parallel graphics processors to provide higher throughput for compute-intensive workloads and achieving significant performance gains without the hidden cost of scale-out architecture.

iRender is currently providing GPU Cloud for AI/DL service for users training their models. With our high configuration and performance machines, you can install any software you need for your demands. Just a few clicks, you are able to get access to our machine and take full control of it. Your model training will speed up 10 times or even 50 times faster.

For more information, please contact us or sign up here and try using our services!

Source: h2o.ai, nvidia.com