DeepAI AI Chat
Log In Sign Up

A backward pass through a CNN using a generative model of its activations

by   Huayan Wang, et al.

Neural networks have shown to be a practical way of building a very complex mapping between a pre-specified input space and output space. For example, a convolutional neural network (CNN) mapping an image into one of a thousand object labels is approaching human performance in this particular task. However the mapping (neural network) does not automatically lend itself to other forms of queries, for example, to detect/reconstruct object instances, to enforce top-down signal on ambiguous inputs, or to recover object instances from occlusion. One way to address these queries is a backward pass through the network that fuses top-down and bottom-up information. In this paper, we show a way of building such a backward pass by defining a generative model of the neural network's activations. Approximate inference of the model would naturally take the form of a backward pass through the CNN layers, and it addresses the aforementioned queries in a unified framework.


page 6

page 7

page 8

page 10

page 11


Answering Hindsight Queries with Lifted Dynamic Junction Trees

The lifted dynamic junction tree algorithm (LDJT) efficiently answers fi...

DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients

We propose DoReFa-Net, a method to train convolutional neural networks t...

Reversible designs for extreme memory cost reduction of CNN training

Training Convolutional Neural Networks (CNN) is a resource intensive tas...

Sample-Specific Output Constraints for Neural Networks

Neural networks reach state-of-the-art performance in a variety of learn...

Neural Networks with Recurrent Generative Feedback

Neural networks are vulnerable to input perturbations such as additive n...

Neural Nets via Forward State Transformation and Backward Loss Transformation

This article studies (multilayer perceptron) neural networks with an emp...

Deep learning: Technical introduction

This note presents in a technical though hopefully pedagogical way the t...