5 papers · 1 filter
Thinking Deeper, Not Longer: Memory-Efficient Test-Time Reasoning with Depth-Recurrent Transformers for Compositional Generalization
Hung-Hsuan Chen
Standard Transformers have a fixed computational depth, limiting their ability to generalize to tasks that require variable-depth reasoning. The usual remedy, Chain-of-Thought (CoT…
SCPL: Enhancing Neural Network Training Throughput with Decoupled Local Losses and Model Parallelism
Ming-Yao Ho, Cheng-Kai Wang, You-Teng Lin +1
Adopting large-scale AI models in enterprise information systems is often hindered by high training costs and long development cycles, posing a significant managerial challenge. Th…
Contrastive ECOC: Learning Output Codes for Adversarial Defense
Che-Yu Chou, Hung-Hsuan Chen
Although one-hot encoding is commonly used for multiclass classification, it is not always the most effective encoding mechanism. Error Correcting Output Codes (ECOC) address multi…
DeInfoReg: A Decoupled Learning Framework for Better Training Throughput
Zih-Hao Huang, You-Teng Lin, Hung-Hsuan Chen
This paper introduces Decoupled Supervised Learning with Information Regularization (DeInfoReg), a novel approach that transforms a long gradient flow into multiple shorter ones, t…
TTSWING: a Dataset for Table Tennis Swing Analysis
Che-Yu Chou, Zheng-Hao Chen, Yung-Hoh Sheu +2
We introduce TTSWING, a novel dataset designed for table tennis swing analysis. This dataset comprises comprehensive swing information obtained through 9-axis sensors integrated in…