Pith. Sign in

RECORD

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

PaperPDF from arXiv

data files

Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
Huayan Wang, Anna Chen, Yi Liu, Dileep George, D. Scott Phoenix

Computer Vision and Pattern Recognition · Submitted 8 Nov 2016

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim In this paper, we show a way of building such a backward pass by defining a generative model of the neural network's activations.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID1611.02767 (version 1)
Pith IDQ2FXXASX, Pith's permanent record of this paper and its review
CitationAnna Chen, Dileep George, D. Scott Phoenix, Huayan Wang, Yi Liu. "A backward pass through a CNN using a generative model of its activations." arXiv:1611.02767 (2016). https://pith.science/paper/Q2FXXASX
Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 4,957 input tokens.

What the abstract itself states, as probabilities: claims first 0.06 · states limitation 0.20 · human participants 0.03 · controlled comparison 0.03 · randomized assignment 0.01 · states quantitative result 0.06 · compares with prior results 0.10 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; .

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.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 Neural networks have shown to be a practical way of building a very complex mapping between a pre-specified input space and output space. background 1.00 0.08 0.03 asserted 0.98
s1 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. background 1.00 0.08 0.03 asserted 0.94
s2 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. gap 0.83 0.10 0.03 asserted 0.99
s3 One way to address these queries is a backward pass through the network that fuses top-down and bottom-up information. background 0.78 0.07 0.07 asserted 0.67
s4 In this paper, we show a way of building such a backward pass by defining a generative model of the neural network's activations. contribution 0.82 0.92 0.60 asserted 1.00
s5 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. interpretation 0.38 0.67 0.53 asserted 0.81

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from argument 0.62.

What was examined, and what was not

sources read
  • arXiv listing abstract v1
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper
Original abstract

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.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

What this paper does with the arXiv papers it cites

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

What it cites · 5 sentences without a confident label

Uses the method or tool of 1 paper
Compares with 3 papers
Cites as support 1 paper
  • Notably, as we have observed from Fig. [ref] and Fig. [ref], and also demonstrated by other researchers [ref] [ref] , contemporary CNNs retain most of the visual attributes at its top layers (and refactor them in a way that is amenable to the classification task).

    Inverting Visual Representations with Convolutional Networks · Discussions and Future Work · 0.99

Cites as background 5 papers
  • There are other forms of backward pass such as deconv-net [ref] , guided-backprop [ref] , auto-encoders [ref] [ref] [ref] , hierarchical rectified Gaussians [ref] , and inverting representations [ref] .

    Auto-Encoding Variational Bayes · Introduction · 0.93

  • There are other forms of backward pass such as deconv-net [ref] , guided-backprop [ref] , auto-encoders [ref] [ref] [ref] , hierarchical rectified Gaussians [ref] , and inverting representations [ref] .

    Striving for Simplicity: The All Convolutional Net · Introduction · 0.93

  • There are other forms of backward pass such as deconv-net [ref] , guided-backprop [ref] , auto-encoders [ref] [ref] [ref] , hierarchical rectified Gaussians [ref] , and inverting representations [ref] .

    Stacked What-Where Auto-encoders · Introduction · 0.93

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.