18537588. MULTIMODAL UNSUPERVISED META-LEARNING METHOD AND APPARATUS simplified abstract (ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE)
MULTIMODAL UNSUPERVISED META-LEARNING METHOD AND APPARATUS
Organization Name
ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE
Inventor(s)
Jeong-Min Yang of Daejeon (KR)
Byung-Hyun Yoo of Daejeon (KR)
MULTIMODAL UNSUPERVISED META-LEARNING METHOD AND APPARATUS - A simplified explanation of the abstract
This abstract first appeared for US patent application 18537588 titled 'MULTIMODAL UNSUPERVISED META-LEARNING METHOD AND APPARATUS
Simplified Explanation
This patent application describes a method and apparatus for multimodal unsupervised meta-learning, which involves training an encoder to extract features from single-modal signals in a multimodal dataset, generating tasks based on these features, deriving learning methods from these tasks, and training a model to perform target tasks based on the extracted features.
- The method involves training an encoder to extract features from individual single-modal signals in a multimodal dataset.
- Tasks are generated based on these features, and learning methods are derived from these tasks using the encoder.
- A model is then trained to perform target tasks using the learning method and features extracted from a small number of target datasets.
Key Features and Innovation
- Training an encoder to extract features from single-modal signals in a multimodal dataset.
- Generating tasks based on these features.
- Deriving learning methods from the tasks using the encoder.
- Training a model to perform target tasks based on the extracted features.
Potential Applications
This technology can be applied in various fields such as computer vision, natural language processing, and speech recognition for improved performance in multimodal tasks.
Problems Solved
This technology addresses the challenge of effectively learning from multimodal datasets without the need for labeled data, leading to better performance in target tasks.
Benefits
- Improved performance in target tasks.
- Reduced reliance on labeled data for training models.
- Enhanced efficiency in learning from multimodal datasets.
Commercial Applications
The technology can be utilized in industries such as healthcare, finance, and autonomous vehicles for tasks that require processing multiple types of data simultaneously.
Prior Art
Researchers can explore prior work in the fields of unsupervised learning, meta-learning, and multimodal data processing to understand the existing knowledge in this area.
Frequently Updated Research
Researchers are constantly exploring new methods and techniques in multimodal unsupervised meta-learning to enhance the capabilities and applications of this technology.
Questions about Multimodal Unsupervised Meta-Learning
1. What are the potential limitations of applying multimodal unsupervised meta-learning in real-world scenarios? 2. How does this technology compare to traditional supervised learning methods in terms of performance and efficiency?
Original Abstract Submitted
Disclosed herein are a multimodal unsupervised meta-learning method and apparatus. The multimodal unsupervised meta-learning method includes training, by a multimodal unsupervised feature representation learning unit, an encoder configured to extract features of individual single-modal signals from a source multimodal dataset, generating, by a multimodal unsupervised task generation unit, a source task based on the features of individual single-modal signals, deriving, by a multimodal unsupervised learning method derivation unit, a learning method from the source task using the encoder, and training, by a target task performance unit, a model based on the learning method and features extracted from a small number of target datasets by the encoder, thus performing the target task.
