DLID: Deep Learning for Domain Adaptation by Interpolating between Domains

By Sumit Chopra, Suhrid Balakrishnan, Raghuraman Gopalan. 2013

207 citations recorded by the source corpus.

Subfields: Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Representation & Self-Supervised Learning

Citations in the corpus

DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition (2013, methodology): “We evaluate on the SUN-397 scene dataset, as well as datasets used to evaluate domain adaptation performance directly (Chopra et al., 2013; Kulis et al., 2011).”
Domain-Adversarial Training of Neural Networks (2015, methodology): “Our training protocol is adopted from Gong et al. (2013); Chopra et al. (2013); Long and Wang (2015) as during adaptation we use all available labeled source examples and unlabeled target examples (the premise of our method is the abundance of unlabeled data in the target domain).”
Cross Modal Distillation for Supervision Transfer (2015, methodology): “d sparsely labeled target domains, [1] use the source models as a parameter regularizer for target models, [8, 20] combine these two approaches into a single joint optimization problem. Chopra et al. [4] introduced one of the ﬁrst deep architectures for visual adaptation by replicating feature extra”
Deep CORAL: Correlation Alignment for Deep Domain Adaptation (2016, background): “DLID [15] trains a joint source and target CNN architecture with two adaptation layers.”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “ image sets [215, 212]. [109] proposes online adaptation of image classiﬁer to user generated content in social computing applications. Furthermore, as discussed in Section 4, ﬁne-tuning a deep model [132, 128, 133, 134, 135, 21], pretrained on ImageNet (source), for a new dataset (target), can be s”
Return of Frustratingly Easy Domain Adaptation (2015, background): “DLID (Chopra, Balakrishnan, and Gopalan 2013) trains a joint source and target CNN architecture with an “interpolating path” between the source and target domain.”
A Deeper Look at Dataset Bias (2015, background): “tions. A large number of studies have recently proposed improvements over the CNN architecture of Krizhevsky et al. [26] with the aim to better suit an ever increasing typology of visual applications [21,45,35]. At the same time, the activation values of the ﬁnal hidden layers have quickly gained th”
SpotTune: Transfer Learning Through Adaptive Fine-Tuning (2018, background): “Recently, transfer learning based on deep neural networks has received significant attention in the community [12, 6, 7, 24, 13].”
Central Moment Discrepancy (CMD) for Domain-Invariant Representation Learning (2017, methodology): “Using this ”fully-transductive” protocol, we compare our method with other state-of-the-art approaches including DLID (Chopra et al., 2013), DDC (Tzeng et al., 2014), DAN (Long et al., 2015), Deep CORAL (Sun & Saenko, 2016), and DANN (Ganin et al., 2016), based on fine-tuning of the baseline model A”
Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation (2016, methodology): “Accuracy (mean ± std %) on the Office dataset with the standard unsupervised domain adaptation protocol used in [7,39].”
Revisiting Batch Normalization For Practical Domain Adaptation (2016, background): “BN layer has become a standard component in recent top-performing CNN architectures, such as deep residual network [25], and Inception V3 [26].”
DASH-N: Joint Hierarchical Domain Adaptation and Feature Learning (2015, methodology): “It can be seen from Tables II and III that our results, even using only features extracted from the first layer, are consistently better than that of other algorithms in all the domain settings, except the results obtained by using features extracted from deep networks [41], [42].”
Scatter Component Analysis: A Unified Framework for Domain Adaptation and Domain Generalization (2015, background): “Algorithms based on hierarchical non-linear features or deep learning are also capable of producing powerful domain adaptive features [14], [52], [53], [54], [55], [56].”
Learning Semantic Image Representations at a Large Scale (2014, methodology): “3: Deep features dramatically outperforms the baseline SURF feature available with the Office dataset as well as the deep adaptive method of [16].”