Downsampling leads to Image Memorization in Convolutional Autoencoders

Memorization of data in deep neural networks has become a subject of significant research interest. In this paper, we link memorization of images in deep convolutional autoencoders to downsampling through strided convolution. To analyze this mechanism in a simpler setting, we train linear convolutional autoencoders and show that linear combinations of training data are stored as eigenvectors in the linear operator corresponding to the network when downsampling is used. On the other hand, networks without downsampling do not memorize training data. We provide further evidence that the same effect happens in nonlinear networks. Moreover, downsampling in nonlinear networks causes the model to not only memorize linear combinations of images, but individual training images. Since convolutional autoencoder components are building blocks of deep convolutional networks, we envision that our findings will shed light on the important phenomenon of memorization in over-parameterized deep networks.

READ FULL TEXT

page 4

page 10

page 18

page 19

page 20

research
10/06/2017

Deep Convolutional Neural Networks as Generic Feature Extractors

Recognizing objects in natural images is an intricate problem involving ...
research
02/20/2018

i-RevNet: Deep Invertible Networks

It is widely believed that the success of deep convolutional networks is...
research
02/21/2019

Convolutional Analysis Operator Learning: Dependence on Training Data

Convolutional analysis operator learning (CAOL) enables the unsupervised...
research
12/31/2018

Soft-Autoencoder and Its Wavelet Shrinkage Interpretation

Deep learning is a main focus of artificial intelligence and has greatly...
research
02/12/2019

Manifestation of Image Contrast in Deep Networks

Contrast is subject to dramatic changes across the visual field, dependi...
research
06/02/2018

Autoencoders Learn Generative Linear Models

Recent progress in learning theory has led to the emergence of provable ...
research
04/12/2021

Epigenetic evolution of deep convolutional models

In this study, we build upon a previously proposed neuroevolution framew...

Please sign up or login with your details

Forgot password? Click here to reset