A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Filters
Architecture Disentanglement for Deep Neural Networks
[article]
2021
arXiv
pre-print
Preserved Fulltext
Other Versions
Understanding the inner workings of deep neural networks (DNNs) is essential to provide trustworthy artificial intelligence techniques for practical applications. ...
In this paper, we introduce neural architecture disentanglement (NAD) to fill the gap. ...
Introduction A fundamental problem in using deep neural networks (DNNs) is our inability to understand their inner workings, which is crucial in many real-world applications, including healthcare, criminal ...
arXiv:2003.13268v2
fatcat:ge33yia6vvg7pcf7fawv2lrina
Differentiable Disentanglement Filter: an Application Agnostic Core Concept Discovery Probe
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2019; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Meanwhile disentangling the actual core concepts engrained in the word embeddings (like word2vec or BERT) or deep convolutional image recognition neural networks (like PG-GAN) is difficult and some success ...
In this paper we propose a novel neural network nonlinearity named Differentiable Disentanglement Filter (DDF) which can be transparently inserted into any existing neural network layer to automatically ...
Successful disentanglement of the core concepts in each layer is only the first step towards understanding the internal logic of the deep neural network. ...
arXiv:1907.07507v2
fatcat:a275ltqghjg2bpkmm254cjstwm
A Deeper Look at the Unsupervised Learning of Disentangled Representations in β-VAE from the Perspective of Core Object Recognition
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Artificial Neural Networks, computational graphs consisting of weighted edges and mathematical operations at vertices, are loosely inspired by neural networks in the brain and have proven effective at ...
(Pinto et al., 2008) (DiCarlo et al., 2012) For many data analysis tasks, learning representations where each dimension is statistically independent and thus disentangled from the others is useful. ...
(Geron, 2017) Human Performance vs Deep Neural Networks With our understanding of core object recognition in the brain and of general deep neural network architectures, it is useful to examine explicit ...
arXiv:2005.07114v1
fatcat:6754eedw2nbexawoblecfbuuda
Joint Parameter Discovery and Generative Modeling of Dynamic Systems
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
The neural framework uses a deep latent variable model to disentangle the system physical parameters from canonical coordinate observations. ...
We present a neural framework for estimating physical parameters in a manner consistent with the underlying physics. ...
We propose a neural framework for join parameter discovery and generative modeling of Hamiltonian systems that merges deep latent variable models and physical constraint embedded neural networks, 2. ...
arXiv:2103.10905v1
fatcat:tbdxwzr7lbdljcyrnml6pvlnvi
Multi-input Architecture and Disentangled Representation Learning for Multi-dimensional Modeling of Music Similarity
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
To achieve this, we propose a multi-input deep neural network architecture, which simultaneously processes mel-spectrogram, CENS-chromagram and tempogram in order to extract informative features for the ...
Current representation learning strategies pursue the disentanglement of such factors from deep representations, resulting in highly interpretable models. ...
Acknowledgements This research was partially supported by H2020 EU project AI4Media -A European Excellence Centre for Media, Society and Democracy -under Grand Agreement 951911. ...
arXiv:2111.01710v1
fatcat:lq33o4ka3jhzfhoaxuhvrdzgbq
Differentiable Frequency-based Disentanglement for Aerial Video Action Recognition
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
We use this differentiable mask prior to enable the neural network to intrinsically learn disentangled feature representations via an identity loss function. ...
Our formulation empowers the network to inherently compute disentangled salient features within its layers. ...
Deep neural networks for human activity recognition [2] , [3] have been widely used for ground-camera scenes [4] , [5] , where the human actors take a high fraction of the pixels in the video scenes ...
arXiv:2209.09194v2
fatcat:3llvgn4ewbbkvenuc7tyqt7vpa
Knowledge Consistency between Neural Networks and Beyond
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
This paper aims to analyze knowledge consistency between pre-trained deep neural networks. ...
We propose a generic definition for knowledge consistency between neural networks at different fuzziness levels. ...
INTRODUCTION Deep neural networks (DNNs) have shown promise in many tasks of artificial intelligence. ...
arXiv:1908.01581v2
fatcat:3d236kykxbav3gjco6qlanz25q
Unsupervised deep learning identifies semantic disentanglement in single inferotemporal neurons
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Deep supervised neural networks trained to classify objects have emerged as popular models of computation in the primate ventral stream. ...
Unlike deep classifiers, beta-VAE "disentangles" sensory data into interpretable latent factors, such as gender or hair length. ...
Acknowledgments We would like to thank Raia Hadsell, Zeb Kurth-Nelson and Koray Kavukcouglu for comments on the manuscript. ...
arXiv:2006.14304v1
fatcat:xt2cluzicbbgbc2quo2oeofmj4
An Explicit Local and Global Representation Disentanglement Framework with Applications in Deep Clustering and Unsupervised Object Detection
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
The code for our experiments is at https://github.com/51616/split-vae . ...
Finally, we establish connections between SPLIT and recent research in cognitive neuroscience regarding the disentanglement in human visual perception. ...
The authors would like to thank Murray Shanahan for discussions on the preliminary ideas and results, which built the foundation for this work. ...
arXiv:2001.08957v2
fatcat:2qjbomhn5zdntdojaq5y4jivqe
Neural Network Renormalization Group
2018
Physical Review Letters
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2018 pre-print arXiv:1802.02840v3 fatcat:sxngikae7vhy5bvwhpq2i5fm3aOther Versions
We present a variational renormalization group (RG) approach using a deep generative model based on normalizing flows. ...
Conversely, the neural net directly maps independent Gaussian noises to physical configurations following the inverse RG flow. ...
Figure 2 . 2 (a) A reference neural network architecture with only disentanglers. ...
doi:10.1103/physrevlett.121.260601
fatcat:4kolmbywu5cnrk2aipge2kh5qa
A Short Survey of Systematic Generalization
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
We hope this paper provides a background and is beneficial for discoveries in future work. ...
Acknowledgments We thank Liang Zhao for beneficial discussions, suggestions, and adding information. We also thank Yi Yang for the helpful advice. ...
Transformers (Vaswani et al. 2017 ) are modern neural network architectures with self-attention. ...
arXiv:2211.11956v1
fatcat:n5kdiovowjdplpc6olyjgpbjna
Disentangling Physical Dynamics From Unknown Factors for Unsupervised Video Prediction
2020
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Since physics is too restrictive for describing the full visual content of generic videos, we introduce PhyDNet, a two-branch deep architecture, which explicitly disentangles PDE dynamics from unknown ...
Ablation studies also highlight the important gain brought out by both disentanglement and PDE-constrained prediction. ...
Deep video prediction Deep neural networks have recently achieved state-of-the-art performances for datadriven video prediction. ...
doi:10.1109/cvpr42600.2020.01149
dblp:conf/cvpr/GuenT20a
fatcat:uqz3dewzrfhb7fr5y6o5xge3aq
Adversarial Learning of Deepfakes in Accounting
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
In this work, we show an adversarial attack against CAATs using deep neural networks. ...
Second, we show that adversarial autoencoder neural networks are capable of learning a human interpretable model of journal entries that disentangles the entries latent generative factors. ...
Acknowledgements We thank the members of the statistics department at Deutsche Bundesbank for their valuable review and remarks. ...
arXiv:1910.03810v1
fatcat:7lzr63ahcbh7vm4n2y4xxr33bq
Disentangling Physical Dynamics from Unknown Factors for Unsupervised Video Prediction
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Since physics is too restrictive for describing the full visual content of generic videos, we introduce PhyDNet, a two-branch deep architecture, which explicitly disentangles PDE dynamics from unknown ...
Ablation studies also highlight the important gain brought out by both disentanglement and PDE-constrained prediction. ...
Deep video prediction Deep neural networks have recently achieved state-of-the-art performances for datadriven video prediction. ...
arXiv:2003.01460v2
fatcat:fviqde6rczf4hjxxnb2tzf53uq
Analyzing and Reducing the Damage of Dataset Bias to Face Recognition With Synthetic Data
2019
2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Furthermore, our study shows that current neural network architectures cannot disentangle face pose and facial identity, which limits their generalization ability. 2) We pre-train neural networks with ...
Our analysis reveals that deeper neural network architectures can generalize better to unseen face poses. ...
Deep networks cannot disentangle face pose from facial identity. ...
doi:10.1109/cvprw.2019.00279
dblp:conf/cvpr/KortylewskiESGM19
fatcat:iuzo4xqjrfgrtoe74tbso7qesq
« Previous
Showing results 1 — 15 out of 15,023 results