paper-with-me

Papers

Head2Toe: Utilizing Intermediate Representations for Better Transfer Learning

2022-01-10 · Utku Evci, Vincent Dumoulin, Hugo Larochelle, Michael C. Mozer

Transfer-learning methods aim to improve performance in a data-scarce target domain using a model pretrained on a data-rich source domain. A cost-efficient strategy, linear probing, involves freezing the source model and training a new classification head for the target domain. This strategy is outperformed by a more costly but state-of-the-art method -- fine-tuning all parameters of the source model to the target domain -- possibly because fine-tuning allows the model to leverage useful information from intermediate layers which is otherwise discarded by the later pretrained layers. We explore the hypothesis that these intermediate layers might be directly exploited. We propose a method, Head-to-Toe probing (Head2Toe), that selects features from all layers of the source model to train a classification head for the target-domain. In evaluations on the VTAB-1k, Head2Toe matches performance obtained with fine-tuning on average while reducing training and storage cost hundred folds or more, but critically, for out-of-distribution transfer, Head2Toe outperforms fine-tuning.

📄 PDF Abstract BibTeX arXiv:2201.03529

Code (2)

google-research/head2toe 공식 구현 tf
arkel23/clca pytorch

Tasks

Transfer Learning

Similar Papers 제목 키워드 기반

Head2Toe: Utilizing Intermediate Representations for Better OOD Generalization

2021-09-29 · Utku Evci, Vincent Dumoulin, Hugo Larochelle, Michael Curtis Mozer

Transfer-learning methods aim to improve performance in a data-scarce target domain using a model pretrained on a data-rich source domain. A cost-efficient strategy, linear probing, involves freezing the source model an…

Transfer Learning

SHERL: Synthesizing High Accuracy and Efficient Memory for Resource-Limited Transfer Learning

2024-07-10 · Haiwen Diao, Bo Wan, Xu Jia, Yunzhi Zhuge 외

Parameter-efficient transfer learning (PETL) has emerged as a flourishing research field for adapting large pre-trained models to downstream tasks, greatly reducing trainable parameters while grappling with memory challe…

Transfer Learning

KTAN: Knowledge Transfer Adversarial Network

2018-10-18 · Peiye Liu, Wu Liu, Huadong Ma, Tao Mei 외

To reduce the large computation and storage cost of a deep convolutional neural network, the knowledge distillation based methods have pioneered to transfer the generalization ability of a large (teacher) deep network to…

image-classificationImage ClassificationKnowledge Distillationobject-detection+2

Exploring the Role of Task Transferability in Large-Scale Multi-Task Learning

2022-04-23 · NAACL 2022 7 · Vishakh Padmakumar, Leonard Lausen, Miguel Ballesteros, Sheng Zha 외

Recent work has found that multi-task training with a large number of diverse tasks can uniformly improve downstream performance on unseen target tasks. In contrast, literature on task transferability has established tha…

Multi-Task LearningRepresentation Learning

UHKD: A Unified Framework for Heterogeneous Knowledge Distillation via Frequency-Domain Representations

2025-10-28 · Fengming Yu, Haiwei Pan, Kejia Zhang, Jian Guan 외 arxiv

Knowledge distillation (KD) is an effective model compression technique that transfers knowledge from a high-performance teacher to a lightweight student, reducing computational and storage costs while maintaining compet…

Knowledge DistillationModel Compression