Kazunori D. Yamada

This study develops a potent B0AT1 inhibitor effective in a PKU mouse model by targeting an allosteric site.

This paper introduces an appendable memory system for AI to store and recall information without parameter updates, using probabilized data to avoid learning dataset features.

This paper introduces a new mechanism called 'relation' that processes context information efficiently, achieving similar performance to attention mechanisms with lower time complexity.

This study predicts and characterizes disorder-order transition regions in proteins using deep learning methods, finding that most transitions prefer ordered states over disordered ones.

This paper develops a novel RNN architecture, YamRNN, with fewer parameters and faster learning performance compared to LSTM and GRU on sequence classification tasks.

This paper introduces YamAdam, an adaptive optimizer with no user-specified hyperparameters, and shows it outperforms 30 other optimizer settings across multiple tasks.

The paper introduces a de novo amino acid sequence profile generator using LSTM, showing improved performance over existing methods for predicting beta-strands.

This paper introduces a novel scoring function for profile alignment using a derivative-free neural network, improving alignment sensitivity and precision for remote sequence pairs.