Overview / Usage

oneAPI an open, cross-architecture programming model gives wings to the developers to use a single code base across multiple architectures to achieve accelerated computing. The Intel oneAPI Base Toolkit contains a core set of tools and libraries for developing high-performance, data-centric applications across diverse architectures. In short it offers a powerful performance-based programming ecosystem.

It is becoming more obvious that the future of computing is not in a single chip for everything, nevertheless in several chips for several things. The consumers can choose, combine, and match from an ocean of accelerators for their explicit requirements. The objective is to aid developers having one set of code to be run seamlessly for all the chips.

Adapting such unified progresses like oneAPI offers flexibility and can get rid of the need to toil with diverse codebases, tools as well as programming languages. It augurs to simplify the software development and delivers relentless performance for accelerated computing without proprietary lock-in, while also assisting the integration of existing code. With oneAPI, developers can choose the best accelerator architecture for the specific problem they are vying to solve without the need to rewrite software for the ensuing architecture/platform.

Machine learning has been growing tremendously in the market and has vast application possibilities and opportunities. Learning ML has become almost inevitable for engineers to find best results and increased productivity. Many tools and software packages are available to enable the ML learning easier. oneAPI from Intel has been a boon in the market and many applications are being developed with the same. The book explores the ML algorithms, concepts and implementation with relevant theoretical explanation and practical implementation with latest and trendy tools which include oneAPI. The contents are curated in such a way that it caters to the needs of everyone from novice to expert level.

The installations and practices to get used to the Intel DevCloud, Jupyter notebook and then the machine learning workflow which is one of the most relevant workloads nowadays are provided. The visualization tools, the classification, regression, bagging and boosting algorithms along with their relevant implementation codes offers an enjoyable learning experience while also demonstrating the solid optimized performance of oneAPI. The classification problems are taken up to make the readers understand the power of optimization with the additional details on Intel tools provided for an enhanced development experience addressing the varying demands of the readers.

GitHub links for the codes have been presented for easier accessibility.

Let’s unleash the true power of our code across diverse architectures!

Methodology / Approach

1. Intel oneAPI – An introductory discussion.

a. Why oneAPI?

b. What is there for us with oneAPI?

c. Features and learning resources.

1. Intel oneAPI – Toolkits – An exploratory analysis.

a. Intel oneAPI toolkits

b. Details in brief about all the tool kits.

c. References and learning materials.

1. Intel DevCloud – Get everything onto cloud.

a. Power of DevCloud

b. Registration process

c. Jupyter notebook and DevCloud

d. DevCloud Commands

1. What is Machine Learning? – Introduction.

a. Types of Machine Learning with examples

b. The ML framework

c. Deep Learning vs. Machine Learning.

d. Where to use Machine Learning and where Deep Learning?

1. The Tools and Pre-requisites <>

a. The under- and over-fitting, regularization, and cross-validation.

b. Intel Extension for Scikit-learn

c. Examples and usage.

1. Supervised Learning

a. Introduction to Supervised Learning

b. What is regression?

c. Where is regression useful?

d. Steps in regression.

e. Classification problems.

f. K-Nearest Neighbors Walk through and implementation

g. Linear vs. Logistic Regression

h. The metrics - cost functions, regularization, feature selection, and hyper-parameters

i. The importance of bias and variance.

1. Support Vector Machines <>

a. What is SVM?

b. How does it work?

c. Implementation and testing

d. Cost functions of SVM

1. Decision Trees

a. What are Decision Trees all about?

b. Classification problem with decision trees.

c. Random Forest Classifier or Decision Trees?

d. Implementation.

1. Bagging

a. Why is bagging important?

b. Variance and Bagging – What’s the connect?

c. Implementation

1. Boosting and Stacking

a. Variance and Boosting – What is the connect?

1. Unsupervised Learning techniques

a. Clustering techniques. <>

b. Dimensionality reduction. <>

1. Advanced AI Tool Kits.

a. Intel AI Analytics Tool Kit

b. Intel OpenVINO

c. Intel Distribution for Python

d. Intel Machine Learning Tool Kits.

Case studies.

Technologies Used

Intel oneAPI

Repository

https://github.com/shriramkv/MachineLearningwithoneAPI