Python-tensorflow-oneDNN

Predicting Consumer Sentiment.

To build the predictive model, participants will be required to use oneAPI libraries such as oneDNN (Deep Neural Network Library), which will enable efficient training and inference on a variety of hardware platforms.

Dataset Preparation: Collect a dataset of images for a particular classification task, such as recognizing different types of flowers.

Data Preprocessing: Preprocess the images by resizing them to a fixed size and normalizing the pixel values.

Model Selection: Choose a deep neural network model architecture that is suitable for the classification task, such as a Convolutional Neural Network (CNN).

Model Training: Use OneDNN to train the model on the preprocessed dataset. Using Python and the Intel® oneAPI AI Analytics Toolkit to train a machine learning model on a preprocessed dataset using the oneDNN library.

Predicting Consumer Sentiment Problem Statement: Businesses can use consumer sentiment to know their customers and improve their sales by formulating a strategy adopting the consumer reactions towards the different types of content . By understanding consumer sentiment, they can tailor their messaging and advertising to appeal to their target audience. In this hackathon, participants are tasked with building a predictive model that can analyze social media activity to determine consumer sentiment towards a particular content. The model should take into account various features of the social media data, such as the latest news, resources, topic of the media resources, popularity on the social media. Task: The goal of this challenge is to develop a machine learning model that can predict consumer sentiment towards the different types of content based on social media activity and optimize it using oneAPI libraries To build the predictive model, participants can use Intel® Optimized Frameworks and Intel® oneAPI AI Analytics Toolkit and Libraries such as oneDNN(Deep Neural Network Library), oneMKL(Math Kernel Library) etc. which will enable efficient training and inference on a variety of hardware platforms. Note: It is mandatory to use the Intel® oneAPI AI Analytics Toolkit and Libraries for the solution. Leverage the libraries and optimizations for the maximum efficiency and better models. You can explore the Intel® oneAPI AI Analytics Toolkit and Libraries to find the related libraries and optimizations for improving performance which will make your solution stand-out. Some of the useful links to explore more about oneAPI are mentioned below.

1. oneAPIDeep Neural Network Library
2. Intel oneAPI Math Kernel Library (oneMKL)
3. Intel® oneAPI Threading Building Blocks
4. Intel® oneAPI Data Analytics Library
5. Intel® oneAPI DPC++ Library
6. Intel® Optimization for TensorFlow*
7. Intel® Optimization for PyTorch*
8. Intel® Distribution for Python*
9. Intel® Extension for Scikit-learn
10. Intel® Neural Compressor
11. Intel® oneAPI AI Reference Kit Data Description: The dataset for this challenge will include a large amount of news items and their reach on social media data from platforms such as LinkedIn, Facebook, and Google+. The purpose is to analyze consumer sentiment based on the different types of news, contents they get or share on social media. Attributes Values IDLink numeric Title string Headline string Source string Topic string Publish-Date timestamp Facebook numeric Google-Plus numeric LinkedIn numeric SentimentTitle numeric SentimentHeadline numeric Dataset: The dataset for the problem is shared below:
12. Training_Data
13. Test_Data
14. Sample_Submission
15. Data_Dictionary Judging Criteria:
16. Usage of Intel® oneAPI AI Analytics Toolkit and Libraries (40%)
17. Completeness (10%)
18. Correctness & Performance (10%)
19. Scalability (5%)
20. Ease of implementation(5%)
21. Benchmarking 15%
22. Optimization 15% Deliverables: (Ideation Phase)
23. Presentation.pptx: This should contain a description of what you have tried to build, what problem you are solving. It should clearly mention the usage of Intel® oneAPI AI Analytics Toolkit and Libraries
24. Participants are required to submit a comprehensive write-up that details their chosen theme, approach to the problem, and the code used to build their solution. This write up should be in the form of a technical article posted on Medium. This write-up will be evaluated by the judges for the functionality and creativity of the submitted solution. Therefore, it is crucial to submit a complete and clear write-up to increase your chances of winning the competition
25. readme.txt: This should contain clear step-by-step instructions on how to build, deployment and usages of the Web Applications.
26. Usage of Intel DevCloud. Link: DevCloud
27. Exploring the oneAPI toolkits/Libraries and its usage in the product is an essential aspect.
28. Products/Projects without oneAPI as the core component will not qualify for the hackathon Note: The source code that's been submitted should be aligned with the project structure that you find on the Intel AI Reference Kit. Link: Intel® oneAPI AI Reference Kit Deliverables: (POC Round)
29. Solution file containing the predicted sentiments (Output file).
30. Code file for producing the solution file (Source Code file or .ipynb file).
31. Share your project on Github with source codes.
32. Video showing the demo functionalities of the Web application over the YouTube/Drive.
33. Presentation.pptx: This should contain the prototype detailed description of what you have tried to build, what problem you are solving and why your Web Application/solution should be considered for the final round. It should clearly mention the usage of Intel® oneAPI AI Analytics Toolkit and Libraries
34. Implementation using oneAPI is mandatory for the POC completion.
35. Exploring the oneAPI toolkits/Libraries and its usage in the product is an essential aspect.
36. Products/Projects without oneAPI as the core component will not qualify for the hackathon Submission Guidelines:
37. We expect a proof of concept/prototype to be built. Architecture and the usage of Intel® oneAPI AI Analytics Toolkit and Libraries has to be clearly presented in the presentation.
38. The code must be made available through GitHub and the same should be presented in the submission page. Ensure access permissions are proper. You can share the solution, challenges faced, learnings, tech-stack used over the GitHub. You can check out this link for reference: Sample Reference
39. The demonstration video can be presented through YouTube/Drive. Ensure you give access to view.
40. We prefer the projects to be submitted in the DevMesh portal. Link: Devmesh Portal Resources: Videos What is oneAPI What is the Intel AI Analytics Toolkit Introduction to oneDnn oneAPI Deep Neural Networks Library Programming Model and Samples oneAPI Video Processing Library Programming Model and Code Samples oneAPI Collective Communications Library oneAPI Video Processing Library Programming Model and Code Samples oneAPI Threading Building Blocks (oneTBB) The oneAPI Math Kernel Library (oneMKL) oneAPI Collective Communications Library | oneCCL oneDPL | oneAPI DPC++ Library Direct Programming with SYCL (SETUP) The Easiest, The Simplest C++ Parallel Library, oneTBB - SpinScoped MutexLock Making banking secure via bio metrics application built using oneAPI and DPC++ based on SYCL/C++ Parallel C++: Concurrent Containers CUDA to SYCL Migration Tool and Method Data Parallel C++ (DPC++) Programming Model YouTube channel from an Intel innovator follow to know more Toolkits & Library AI Reffernce Kit oneAPI Deep Neural Network Library Intel® oneAPI Threading Building Blocks Intel® oneAPI Data Analytics Library Intel® oneAPI Video Processing Library Intel® oneAPI Collective Communications Library Intel® oneAPI DPC++ Library Intel® Optimization for TensorFlow* Intel® Optimization for PyTorch* Intel® Distribution for Python* Intel® Extension for Scikit-learn Intel® Neural Compressor Workshops oneAPI Hands-on Workshop- Let us SYCL* Accelerating Deep Learning with Intel oneAPI oneDNN and oneMKL: A Hands-on workshop Accelerate ML pipelines using Intel Extension for Scikit-learn* and Modin Dataframe