Revolutionizing Machine Learning with Snap ML: Accelerated Training and Faster Results

Dive into how Snap ML is transforming machine learning with its distributed GPU-accelerated library, drastically reducing training times from hours to seconds. Business sectors, such as financial services, benefit from accelerated model training for tasks like credit default prediction and fraud detection. Snap ML empowers data scientists by enabling easy clustering and enhancing linear models' performance, ultimately revolutionizing the field of data science.

• Machine Learning
• Snap ML
• GPU Acceleration
• Accelerated Training
• Data Science

fshe Follow

Uploaded on Dec 05, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

Presentation Transcript

1. PowerAI Snap ML Understanding the Details Chris Eaton (ceaton@ca.ibm.com) Scott Soutter (soutter@us.ibm.com)

2. What data science methods are used at work? Deep Learning is important and growing rapidly, but a Kaggle survey shows data scientists still rely primarily on machine learning algorithms

3. Enterprises rely on linear models for analytics Regulated industries require explainable/interpretable predictions Models are powered largely by text and structured data Well understood tools; clients have deeper skill using Python, Spark, or R Few of these linear models are cluster GPU accelerated, limiting both scale and speed Snap ML changes this to accelerate machine learning

4. Snap ML Distributed GPU-Accelerated Machine Learning Library Snap Machine Learning (ML) Library APIs for Popular ML Frameworks Logistic Regression Linear Regression Support Vector Machines (SVM) More Coming Soon libGLM (C++ / CUDA Optimized Primitive Lib) Distributed Hyper- Parameter Optimization Distributed Training Distributed Training GPU Acceleration Sparse Data Optimization

5. Snap ML: Training time goes from 1.1 hours to 90 seconds Logistic Regression in TensorFlow(CPU-only) vs Snap ML (with GPUs) 80 46x faster than previous record set by Google 1.1 Hours 46x Faster 60 Transparent to data scientists plugs into existing machine learning codes Runtime (Minutes) 40 Generalized linear models: linear regression, logistic regression, support vector machines 20 92 Seconds FAST and EASY to cluster enable machine learning models for GPU and CPU 0 Google CPU-only Snap ML Power + GPU 89 x86 Servers (CPU-only) 4 POWER9 Servers With GPUs

6. Simple to get started Snap ML is available now as a no-charge download from IBM Financial Services Usage Examples: Predict credit default 23x faster than scikit-learn Runs exclusively on the accelerated IBM Power AC922 or S822LC servers Speed up model training for credit card fraud detection: 32x faster than TensorFlow, 12.5x faster than scikit-learn Can accelerate and scale existing models, with little or no modification Predict stock volatility from 10-k textual reports 35x faster than Apache Spark