Source Training

Source training, encompassing techniques like transfer learning and domain adaptation, aims to leverage pre-trained models and data from one domain (source) to improve performance on a related but different domain (target). Current research focuses on mitigating issues like source bias and distribution shifts, employing methods such as contrastive learning, distributionally robust optimization, and novel architectures like Vision Transformers, to enhance model generalization and robustness. These advancements are crucial for improving the efficiency and reliability of machine learning across various applications, from natural language processing and image recognition to industrial processes like battery health prediction and additive manufacturing.

Papers

September 18, 2024

Extracting Memorized Training Data via Decomposition
Ellen Su, Anu Vellore, Amy Chang, Raffaele Mura, Blaine Nelson, Paul Kassianik, Amin Karbasi
Training Data Microbial Decomposition Training Data Privacy Source Training Training Data Extraction

May 10, 2024

Improving Transferable Targeted Adversarial Attack via Normalized Logit Calibration and Truncated Feature Mixing
Juanjuan Weng, Zhiming Luo, Shaozi Li
Task Transferability Adversarial Sample Attack Success Rate Transferable Adversarial Attack Feature Mixing Logit Calibration Source Training

May 5, 2024

Source-Free Domain Adaptation Guided by Vision and Vision-Language Pre-Training
Wenyu Zhang, Li Shen, Chuan-Sheng Foo
Vision Language Vision Paper Source Free Domain Adaptation Pre Trained Feature Extractor Source Training

April 1, 2024

Source-Aware Training Enables Knowledge Attribution in Language Models
Muhammad Khalifa, David Wadden, Emma Strubell, Honglak Lee, Lu Wang, Iz Beltagy, Hao Peng
Large Language Model Language Model Training Data Attribution Source Training Document Identifier

May 17, 2023

Comparison of Transfer Learning based Additive Manufacturing Models via A Case Study
Yifan Tang, M. Rahmani Dehaghani, G. Gary Wang
Transfer Learning Case Study Consistent Comparison Additive Manufacturing Regression Tree Source Training

April 27, 2023

Maximizing Model Generalization for Machine Condition Monitoring with Self-Supervised Learning and Federated Learning
Matthew Russell, Peng Wang
Deep Learning Self Supervised Learning Transfer Learning Machine Condition Monitoring Source Training

April 14, 2023

Frequency Decomposition to Tap the Potential of Single Domain for Generalization
Qingyue Yang, Hongjing Niu, Pengfei Xia, Wei Zhang, Bin Li
Strong Generalization Full Potential Frequency Domain Single Source Domain Generalization Single Domain Source Training Frequency Decomposition

December 5, 2022

Addressing Distribution Shift at Test Time in Pre-trained Language Models
Ayush Singh, John E. Ortega
Data Augmentation Pre Trained Language Model Unsupervised Domain Adaptation Distribution Shift Test Time Adaptation Test Time Source Training

October 27, 2022

COCO-DR: Combating Distribution Shifts in Zero-Shot Dense Retrieval with Contrastive and Distributionally Robust Learning
Yue Yu, Chenyan Xiong, Si Sun, Chao Zhang, Arnold Overwijk
Distribution Shift Dense Retrieval Distributionally Robust Zero Shot Retrieval Zero Shot Model V Coco Sg Zero Shot Dense Retrieval Source Training

September 7, 2022

Transfer Learning and Vision Transformer based State-of-Health prediction of Lithium-Ion Batteries
Pengyu Fu, Liang Chu, Zhuoran Hou, Jincheng Hu, Yanjun Huang, Yuanjian Zhang
Vision Transformer Transfer Learning Lithium Ion Battery Energy Storage State of Health Source Training

Source Training

Papers

Extracting Memorized Training Data via Decomposition

Improving Transferable Targeted Adversarial Attack via Normalized Logit Calibration and Truncated Feature Mixing

Source-Free Domain Adaptation Guided by Vision and Vision-Language Pre-Training

Source-Aware Training Enables Knowledge Attribution in Language Models

Comparison of Transfer Learning based Additive Manufacturing Models via A Case Study

Maximizing Model Generalization for Machine Condition Monitoring with Self-Supervised Learning and Federated Learning

Frequency Decomposition to Tap the Potential of Single Domain for Generalization

Addressing Distribution Shift at Test Time in Pre-trained Language Models

COCO-DR: Combating Distribution Shifts in Zero-Shot Dense Retrieval with Contrastive and Distributionally Robust Learning

Transfer Learning and Vision Transformer based State-of-Health prediction of Lithium-Ion Batteries