Company Filing History:
Years Active: 2021-2022
Title: Hsiang-Fu Yu: Innovator in Multi-Label Classification and Time Series Forecasting
Introduction
Hsiang-Fu Yu is a notable inventor based in San Jose, CA, with a focus on advancements in machine learning and data processing. He holds 3 patents that showcase his expertise in developing innovative solutions for complex problems.
Latest Patents
One of his latest patents is titled "Extreme Multi-Label Classification." This patent presents a contextual label compression framework that utilizes trained sequence-to-sequence models. The framework processes a set of training data, including received queries and related content, to generate sequences of semantic encodings. These sequences are instrumental in training the models to predict queries for instances of content when relevant information is processed. The model outputs a semantic sequence that can be decoded to produce relevant queries for the content instance, aiding in effective content management.
Another significant patent is "High Dimensional Time Series Forecasting." This patent describes large-scale time series forecasting models that leverage deep learning techniques. The models utilize temporal convolution networks and long short-term memory (LSTM) units of recurrent neural networks (RNNs). They are designed to handle extensive datasets with numerous time series. The architecture provides diverse scaling through a scale-free leveling network and manages sparse time-series data using a gating approach. Additionally, a deep temporally regularized matrix factorization approach is employed to leverage correlations between time series during both training and prediction.
Career Highlights
Hsiang-Fu Yu has contributed to prominent companies such as A9.com, Inc. and Amazon Technologies, Inc. His work in these organizations has significantly impacted the fields of machine learning and data analysis.
Collaborations
He has collaborated with notable professionals in the industry, including Inderjit Dhillon and Rajat Sen. Their collective expertise has fostered advancements in the technologies they have worked on together.
Conclusion
Hsiang-Fu Yu's contributions to innovations in multi-label classification and time series forecasting demonstrate his commitment to advancing technology. His patents reflect a deep understanding of machine learning and its applications in real-world scenarios.