Domain Adaptive Neural Networks for Object Recognition

By Muhammad Ghifary, W. Kleijn, Mengjie Zhang. 2014

468 citations recorded by the source corpus.

We propose a simple neural network model to deal with the domain adaptation problem in object recognition. Our model incorporates the Maximum Mean Discrepancy (MMD) measure as a regularization in the supervised learning to reduce the distribution mismatch between the source and target domains in the latent space. From experiments, we demonstrate that the MMD regularization is an effective tool to provide good domain adaptation models on both SURF features and raw image pixels of a particular image data set. We also show that our proposed model, preceded by the denoising auto-encoder pretraining, achieves better performance than recent benchmark models on the same data sets. This work represents the first study of MMD measure in the context of neural networks.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Media Technology, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

In Media Technology

In Representation & Self-Supervised Learning

Citations in the corpus

Learning Transferable Features with Deep Adaptation Networks (2015, background): “A rich line of prior work has focused on learning shallow features by jointly minimizing a distance metric of domain discrepancy (Pan et al., 2011; Long et al., 2013; Baktashmotlagh et al., 2013; Gong et al., 2013; Zhang et al., 2013; Ghifary et al., 2014; Wang & Schneider, 2014).”
Cross Modal Distillation for Supervision Transfer (2015, methodology): “…one of the first deep architectures for visual adaptation by replicating feature extraction for each domain and producing intermediate interpolated domains, while Ghifary et al. [10] showed a single layer neural net could be used to learn the feature transformation between simple domain shifts.”
Simultaneous Deep Transfer Across Domains and Tasks (2015, methodology): “Rather than globally optimizing θD and θrepr, we instead perform iterative updates for the following two objectives given the fixed parameters from the previous iteration: min θD LD(xS , xT , θrepr; θD) (5) min θrepr Lconf(xS , xT , θD; θrepr).”
Deep Domain Confusion: Maximizing for Domain Invariance (2014, neutral): “This effectively learns a domain invariant representation, but again, because the learned network is relatively shallow, it lacks the strong semantic representation that is learned by directly optimizing a classiﬁcation objective with a supervised deep CNN.”
Moment Matching for Multi-Source Domain Adaptation (2018, methodology): “Inspired by the kernel two-sample test [12], Maximum Mean Discrepancy (MMD) is applied to reduce distribution shift in various methods [27, 41, 9, 44].”
Return of Frustratingly Easy Domain Adaptation (2015, methodology): “DANN (Ghifary, Kleijn, and Zhang 2014) incorporates the Maximum Mean Discrepancy (MMD) measure as a regularization to reduce the distribution mismatch.”
Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation (2016, background): “In visual recognition, a fully connected, shallow network pretrained by denoising autoencoders has shown a certain level of effectiveness [34].”
Revisiting Batch Normalization For Practical Domain Adaptation (2016, background): “Ganin et al. [7] devised a gradient reverse layer to reverse the gradient that helps to distinguish the domains of each data sample.”
PixelVAE++: Improved PixelVAE with Discrete Prior (2019, background): “Many machine learning tasks such as domain adaptation [12, 13, 14, 15] and few-shot learning [16, 17, 18] rely on learned representations of observations.”
Co-regularized Alignment for Unsupervised Domain Adaptation (2018, methodology): “Domain alignment has almost become a representative approach for domain adaptation, acting as a crucial component in many recently proposed methods [17, 26, 41, 40, 4, 16, 44, 42, 35].”
A Decade Survey of Transfer Learning (2010–2020) (2020, background): “Firstly, the idea of adaptation layer was proposed by [17].”
A survey: Deep learning for hyperspectral image classification with few labeled samples (2021, methodology): “[90] first proposed a shadow neural network-based domain adaptation model, called DaNN.”
Reducing Domain Gap by Reducing Style Bias (2021, background): “A popular approach is learning a shared feature space across multiple source domains, for example, by minimizing Maximum Mean Discrepancy (MMD) [38, 16] or adversarial feature alignment [31, 32].”
Multi-Modal Domain Adaptation for Fine-Grained Action Recognition (2019, methodology): “D2→ D1 D3→ D1 D1→ D2 D3→ D2 D1→ D3 D2→ D3 Mean MM Source-only 42.5 44.3 42.0 56.3 41.2 46.5 45.5 AdaBN [29] 44.6 47.8 47.0 54.7 40.3 48.8 47.2 MMD [32] 43.1 48.3 46.6 55.2 39.2 48.5 46.8 MCD [45] 42.1 47.9 46.5 52.7 43.5 51.0 47.3 The self-supervision hyper-parameter, λc = 5 was chosen by observing ”
Beyond Sharing Weights for Deep Domain Adaptation (2016, methodology): “In [30], the MMD was used within a shallow neural network architecture.”
A Survey on Negative Transfer (2020, background): “, traditional (statistical) TL [4]–[7], deep TL [8], [9], adversarial TL [10], [11], etc.”
Open Set Domain Adaptation (2017, background): “The MMD distance has been also proposed as regulariser to learn features for source and target samples jointly [14, 46, 29, 30].”
Multi-Modal Domain Adaptation for Fine-Grained Action Recognition (2020, methodology): “Recently, significant attention has been given to deep UDA in other vision tasks [14, 15, 32, 33, 51, 55].”