Duke University’s biomedical engineers have devised an innovative approach to enhance the performance of machine learning models. By combining two machine learning models, one for data collection and the other for analysis, researchers can overcome the technology’s limitations while maintaining accuracy.
The findings are featured in the journal Artificial Intelligence in the Life Sciences.
Conventional machine learning models rely on inputting a dataset for making predictions. However, these tools are constrained by the datasets used for training, which may lack crucial information or introduce bias due to an overabundance of a specific data type.
Instead, researchers have introduced active machine learning, enabling the model to seek additional information or ask questions when data gaps are detected. This questioning ability enhances the model’s accuracy and efficiency compared to passive models.
While active learning is effective, it faces challenges when applied to complex deep neural networks. These models, designed to emulate the human brain, demand extensive data and computing power, limiting their accuracy and effectiveness.
2024-01-20 09:41:03
Original from phys.org
