RESUMEN
Knowledge distillation (KD) is a widely adopted model compression technique for improving the performance of compact student models, by utilizing the "dark knowledge" of a large teacher model. However, previous studies have not adequately investigated the effectiveness of supervision from the teacher model, and overconfident predictions in the student model may degrade its performance. In this work, we propose a novel framework, Teacher-Student Complementary Sample Contrastive Distillation (TSCSCD), that alleviate these challenges. TSCSCD consists of three key components: Contrastive Sample Hardness (CSH), Supervision Signal Correction (SSC), and Student Self-Learning (SSL). Specifically, CSH evaluates the teacher's supervision for each sample by comparing the predictions of two compact models, one distilled from the teacher and the other trained from scratch. SSC corrects weak supervision according to CSH, while SSL employs integrated learning among multi-classifiers to regularize overconfident predictions. Extensive experiments on four real-world datasets demonstrate that TSCSCD outperforms recent state-of-the-art knowledge distillation techniques.
Asunto(s)
Compresión de Datos , Humanos , Conocimiento , Aprendizaje , EstudiantesRESUMEN
This paper focuses on reporting results obtained by the spark plasma sintering (SPS) consolidation and characterization of aluminum-based nanocomposites reinforced with concentrations of 0.5 wt%, 1 wt% and 2 wt% of single-walled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs). Experimental characterization performed by SEM shows uniform carbon nanotube (CNT) dispersion as well as carbon clusters located in the grain boundary of the Al matrix. The structural analysis and crystallite size calculation were performed by X-ray diffraction tests, detecting the characteristic CNT diffraction peak only for the composites reinforced with MWCNTs. Furthermore, a considerable increment in the crystallite size value for those Al samples reinforced and sintered with 1 wt% of CNTs was observed. Hardness tests show an improvement in the composite surface hardness of about 11% and 18% for those samples reinforced with 2 wt% of SWNCTs and MWCNTs, respectively. Conductivity measurements show that the Al samples reinforced with 2 wt% of MWCNTs and with 0.5 wt% SWCNTs reach the highest IACS values of 50% and 34%, respectively.