Corpus ID: 362467

How transferable are features in deep neural networks?

@article{Yosinski2014HowTA,
  title={How transferable are features in deep neural networks?},
  author={Jason Yosinski and Jeff Clune and Yoshua Bengio and Hod Lipson},
  journal={ArXiv},
  year={2014},
  volume={abs/1411.1792}
}
Many deep neural networks trained on natural images exhibit a curious phenomenon in common: on the first layer they learn features similar to Gabor filters and color blobs. [...] Key Result A final surprising result is that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset.Expand
Transferability of features from multiple depths of a deep convolutional neural network
TLDR
This work attempts to quantify how network layers go from general to specific through extracting features from multiple depths and reached the conclusion that the feature from the very last layer of a deep convolutional network are very specific to the source task and using it to learn other classification problems is often sub-optimal. Expand
Neural Dataset Generality
TLDR
A way to analyse and quantify generality among datasets from their accuracies on transferred filters is proposed and it is discovered that particular classes in a dataset themselves are more general than others. Expand
Object recognition using deep convolutional neural networks with complete transfer and partial frozen layers
TLDR
This paper describes an approach of domain transfer, where features learned from a large annotated dataset are transferred to a target domain where less annotated examples are available as is typical for the security and defense domain. Expand
On the generality of neural image features
TLDR
A way to analyse and quantify generality is proposed, found that particular classes in a dataset themselves are more general than others, on several popular character recognition, natural image and medical image datasets. Expand
Deep Networks Can Resemble Human Feed-forward Vision in Invariant Object Recognition
TLDR
This work benchmarked eight state-of-the-art DCNNs, the HMAX model, and a baseline shallow model and compared their results to those of humans with backward masking to demonstrate that shallow nets can outperform deep nets and humans when variations are weak. Expand
Understanding Representations and Reducing their Redundancy in Deep Networks
TLDR
A new regularizer called DeCov is proposed, which leads to significantly reduced overfitting (difference between train and val performance) and greater generalization and takes advantage of the intuition that different features should try to represent different things, an intuition others have explored with similar losses. Expand
Hiding a plane with a pixel: examining shape-bias in CNNs and the benefit of building in biological constraints
TLDR
It is shown that when CNNs are trained end-to-end they learn to classify images based on whatever feature is predictive of a category within the dataset, which raises doubts over the assumption that simply learning end- to-end in standard CNNs leads to the emergence of similar representations to the human visual system. Expand
Learning Transferable Features with Deep Adaptation Networks
TLDR
A new Deep Adaptation Network (DAN) architecture is proposed, which generalizes deep convolutional neural network to the domain adaptation scenario and can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Expand
On the Behavior of Convolutional Nets for Feature Extraction
TLDR
This paper statistically measure the discriminative power of every single feature found within a deep CNN, when used for characterizing every class of 11 datasets, and finds that all CNN features can be used for knowledge representation purposes both by their presence or by their absence. Expand
Neural Networks for Cross-Modal Recognition and Vector Object Generation
TLDR
This report discusses how to learn neural networks more inline with the following two intuitions: (i) the same scene can appear very differently and be depicted in different modalities and (ii) complex objects can be explained by simple primitives. Expand
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 16 REFERENCES
DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition
TLDR
DeCAF, an open-source implementation of deep convolutional activation features, along with all associated network parameters, are released to enable vision researchers to be able to conduct experimentation with deep representations across a range of visual concept learning paradigms. Expand
ImageNet classification with deep convolutional neural networks
TLDR
A large, deep convolutional neural network was trained to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes and employed a recently developed regularization method called "dropout" that proved to be very effective. Expand
Deep Learning of Representations for Unsupervised and Transfer Learning
  • Yoshua Bengio
  • Computer Science
  • ICML Unsupervised and Transfer Learning
  • 2012
TLDR
Why unsupervised pre-training of representations can be useful, and how it can be exploited in the transfer learning scenario, where the authors care about predictions on examples that are not from the same distribution as the training distribution. Expand
Visualizing and Understanding Convolutional Networks
TLDR
A novel visualization technique is introduced that gives insight into the function of intermediate feature layers and the operation of the classifier in large Convolutional Network models, used in a diagnostic role to find model architectures that outperform Krizhevsky et al on the ImageNet classification benchmark. Expand
ICA with Reconstruction Cost for Efficient Overcomplete Feature Learning
TLDR
A robust soft reconstruction cost for ICA is proposed that allows us to learn highly overcomplete sparse features even on unwhitened data and reveals formal connections between ICA and sparse autoencoders, which have previously been observed only empirically. Expand
OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks
TLDR
This integrated framework for using Convolutional Networks for classification, localization and detection is the winner of the localization task of the ImageNet Large Scale Visual Recognition Challenge 2013 and obtained very competitive results for the detection and classifications tasks. Expand
Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations
TLDR
The convolutional deep belief network is presented, a hierarchical generative model which scales to realistic image sizes and is translation-invariant and supports efficient bottom-up and top-down probabilistic inference. Expand
Improving neural networks by preventing co-adaptation of feature detectors
When a large feedforward neural network is trained on a small training set, it typically performs poorly on held-out test data. This "overfitting" is greatly reduced by randomly omitting half of theExpand
What is the best multi-stage architecture for object recognition?
TLDR
It is shown that using non-linearities that include rectification and local contrast normalization is the single most important ingredient for good accuracy on object recognition benchmarks and that two stages of feature extraction yield better accuracy than one. Expand
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
TLDR
This paper proposes a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3%. Expand
...
1
2
...