Thang Doan

Title: Thang Doan: Innovator in Few-Shot Class-Incremental Learning

Introduction

Thang Doan is an accomplished inventor based in San Francisco, CA. He has made significant contributions to the field of machine learning, particularly in few-shot class-incremental learning. His innovative approach combines advanced techniques to enhance the performance of image-text classification tasks.

Latest Patents

Thang Doan holds a patent for a "Minimalist multi-modal approach to few-shot class-incremental learning." This patent describes methods and systems for Few-Shot Class-Incremental Learning (FSCIL) that utilize a combination of Session Specific Prompts (SSP) and hyperbolic distance metrics. These techniques enhance session-wise learning and representation of image-text pairings across differing classes. The methods include a base training session where both text and image features are projected into hyperbolic space for accurate class pairing using a cross-entropy loss function. Subsequent incremental sessions incorporate previously learned SSPs to retain and augment the separability of classes while minimizing the trainable parameters. This innovative approach achieves higher accuracy with fewer trainable parameters compared to traditional models.

Career Highlights

Thang Doan is currently employed at Robert Bosch, where he continues to develop cutting-edge technologies in machine learning. His work focuses on improving the efficiency and effectiveness of classification systems, making significant strides in the field.

Collaborations

Thang collaborates with talented coworkers, including Sima Behpour and Xin Li. Their combined expertise fosters a creative environment that drives innovation and enhances research outcomes.

Conclusion

Thang Doan is a notable inventor whose work in few-shot class-incremental learning is paving the way for advancements in machine learning. His contributions are instrumental in improving image-text classification tasks, showcasing the potential of minimalist approaches in technology.