Open	Boqiang is currently in his 1st year, pursuing a Ph.D. degree in Computer Science at Clemson University. His academic interests include Machine Learning and High-Performance Computing.

Poster Abstract

Neural Networks have emerged as a powerful tool for analyzing complex datasets like images, video, and time-series data, surpassing classical methods in their effectiveness. Leveraging this potential, HPCC Systems offers the Generalized Neural Network Bundle (GNN), which combines the parallel processing capabilities of HPCC Systems with the robust Neural Network functionalities of Keras and TensorFlow.

I upgrade the GNN bundle to utilize the native Tensorflow 2 interface. The upgraded GNN with Tensorflow 2 demonstrated several significant advantages over its previous version:

Latest Tensorflow for Improved Performance: By leveraging Tensorflow 2, the GNN bundle's performance is expected to improve after tests are concluded.

GPU Acceleration for Training: GNN with Tensorflow 2.0 harnesses the power of GPU to train neural networks, enabling efficient and rapid training processes. Performance improvements are expected to be obtained after tests are concluded.

Extensive Pretrained Models: GNN with Tensorflow 2 supports the direct loading of over 70 state-of-the-art neural networks, including ResNet50 and MobileNet. This feature empowers users to integrate these pre-trained models into their AI programs, reducing development time and effort.

Effortless Model Saving and Loading: The upgraded GNN allows users to save and reload their neural network models effortlessly. This functionality facilitates model reuse and sharing among researchers and developers.

Transfer Learning Opportunities: With GNN and Tensorflow 2, users can take advantage of transfer learning by leveraging any pre-trained neural network. Researchers can create innovative solutions for various data analysis tasks by building upon existing knowledge and models.

The combination of HPCC Systems' parallel processing capabilities and the cutting-edge features of Tensorflow 2.0 through the GNN bundle opens up new avenues for efficient and effective data analysis across various domains. Users can harness the power of deep learning and pre-trained models to overcome the limitations of classical data analysis methods and explore novel solutions. The upgraded GNN with Tensorflow 2.0 represents a significant step forward in enabling powerful and scalable analysis of complex datasets, ushering in a new era of advanced data processing and artificial intelligence applications.

Presentation

In this Video Recording, Boqiang provides a tour and explanation of his poster content.

Convert Generalized Neural Network bundle (GNN) to native Tensorflow 2.0

Click on the poster for a larger image.

Open