Pith. sign in

REVIEW 3 major objections 5 minor 50 references

Visualizing Image Content to Explain Novel Image Discovery

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read DEMUD-VIS generates visual explanations of novel image content by inverting reconstruction residuals back into image space, and a user study shows these explanations help people notice within-class novelty.

desk verdict A genuine first step for visual novelty explanations, but the SVD baseline is broken and the explanation/discovery layer mismatch weakens the central claim. read the letter →

arxiv 1908.05006 v2 pith:KR5F2QQN submitted 2019-08-14 cs.LG cs.CVstat.ML

Jake H. Lee , Kiri L. Wagstaff This is my paper · ORCID
classification cs.LGcs.CVstat.ML
keywords noveltydetectionvisualexplanationsimagediscoveryconvolutionalneuralnetworksfeatureinversionreconstructionerrorDEMUDinterpretablemachinelearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces DEMUD-VIS, which it claims is the first method to give visual explanations of novel image content. The pipeline uses a convolutional network to encode images, DEMUD's reconstruction-error ranking to pick images whose content is not yet modeled, and an up-convolutional network to turn the reconstruction and residual feature vectors back into images. The residual image shows what is new in the selected image, while the reconstruction shows what the model already knew. On ImageNet, Mars rover, and stonefly image sets, DEMUD with CNN features, especially from layer fc8, outperformed random and SVD baselines at discovering new classes. A user study found that the explanations shifted judgments toward detecting within-class novelty, although they did not reduce "not sure" responses.

What carries the argument

The central mechanism is the residual vector $r = x' - \hat{x}'$ produced by DEMUD's incremental SVD model, combined with an up-convolutional network (UC DeePSiM) trained to invert CNN feature vectors back into images. The residual is what carries the novelty: it is the part of the image's feature vector that the current model cannot reconstruct. The paper modifies UC DeePSiM to operate on pre-ReLU feature vectors so that the signs in the residual are preserved, and applies a mean-shift transformation to fc6 residuals to make them visualizable. This inversion step is the load-bearing bridge between feature-space novelty and human-comprehensible explanation.

What would settle it

Take a labeled image pair where DEMUD-VIS selects a within-class novelty, mask or inpaint the region highlighted in the residual, and check whether the image's reconstruction error drops to near baseline; if removing the residual-highlighted content does not reduce the novelty score, the visualization is not pointing at the semantically novel content.

Watch

Extended reading notes

Core claim

The paper's central claim is that novelty detection in large image collections becomes useful for discovery only when the algorithm also says what is new, and that for images this explanation can be rendered visually. DEMUD-VIS does this by treating DEMUD's SVD reconstruction and residual in CNN feature space as images to be generated: the reconstruction visualizes content the model already knew, and the residual visualizes the content that fell outside the model. The paper reports near-oracle class discovery on balanced ImageNet, where the fc8 representation discovered a new class on each of the first 13 selections, and best-in-experiment discovery scores on imbalanced and yellow ImageNet subsets, Mars-Curiosity images, and the STONEFLY9 insect data set. In a user study, users given the visual explanations were more likely than users without them to identify within-class novelty for three specific images at p≤0.05, while the overall rate of "yes, new type of object" responses did not change significantly. The paper also documents a trade-off: the best class discovery comes from fc8 features, but the most interpretable visualizations come from fc6 features, so the reported explanations are generated from fc6.

Load-bearing premise

The approach stands on the assumption that the generated residual image faithfully shows the content that made the image novel, and since the authors show the inversion can fail for some network layers, the whole utility rests on that fragile step.

Editorial extensions

If this is right

  • On large image collections, reviewing only the top DEMUD-VIS selections concentrates human attention on genuinely new classes; on balanced ImageNet the first 13 fc8 selections each introduced a new class.
  • The reconstruction-plus-residual explanation pair can expose within-class changes such as dust deposition on a Mars rover instrument or coat-color variation in foxhounds, not just entirely new categories.
  • Mislabeled images can be caught automatically: applied to tiger cub images, the ninth selection was a leopard cub, and the residual highlighted the green leaf and leopard spots that explained the labeling error.
  • Users need some practice with the explanations: in the user study, early selections showed no benefit or even confusion, while later selections showed increased detection of within-class novelty, suggesting that training or reference material would help.
  • Because fc8 gives the best discovery but fc6 gives the best visual interpretability, the current DEMUD-VIS explanations use fc6 features, implying a direct trade-off between discovery quality and explanation quality.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If residual inversion could be made reliable at fc8, the method would combine its best discovery performance with its best explanations; the paper's own mean-shift fix works for fc6 but not fully for fc8, so that is the immediate extension.
  • The approach should transfer to any pretrained CNN whose features can be inverted, so newer architectures with better-trained decoders could improve explanation fidelity without changing the discovery algorithm.
  • The user-study pattern suggests that explanation utility should be measured after user training rather than on first exposure; a testable design would randomize order or provide a tutorial and then measure novelty-judgment accuracy against held-out labels.
  • The residual-as-explanation idea generalizes beyond images: any reconstruction-error detector operating on invertible features could render explanations in the input domain, for example spectra or volumetric medical data, as long as a decoder exists.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes DEMUD-VIS, a pipeline for visual explanations of novel image content. It combines CaffeNet CNN features, the DEMUD reconstruction-error novelty detector, and an up-convolutional network (UC DeePSiM) to invert DEMUD's reconstruction and residual feature vectors into pixel-space images. Experiments cover balanced and imbalanced ImageNet subsets, ImageNet-Yellow, Mars-Curiosity, and STONEFLY9, measuring class discovery nAUC and comparing DEMUD with SVD and random baselines. A user study with 280 complete responses tests whether adding explanations changes human novelty judgments. The paper reports that DEMUD with fc8 features yields the best discovery, while fc6 features yield the most interpretable explanations, and that explanations increase detection of within-class novelty in some cases.

Significance. The problem of explaining unsupervised novelty/class discovery in image data is important and under-served. DEMUD-VIS combines three existing components in a novel way and is, to my knowledge, the first to render DEMUD residuals as images. The paper ships data, features, and scripts, and it includes a user study rather than only anecdotal examples. The explicit discussion of the trade-off between explanation interpretability and detection performance is commendable. However, two load-bearing issues (the degenerate SVD baseline and the fc6/fc8 feature-space mismatch) and the mixed user-study evidence prevent acceptance in the current form.

major comments (3)
  1. [4.2.1, Eq. (5)] The SVD baseline is degenerate as configured. With k set to min(d,n), the SVD model spans the entire data set, so the reconstruction error R(x) in Eq. (5) is identically zero for every item. The paper nonetheless reports SVD nAUC values in Tables 1-3 that are neither zero nor random (e.g., Table 1, balanced ImageNet: SVD CNN-fc8 nAUC70=60.58), which is internally inconsistent. Please specify the actual k used, the tie-breaking rule, or correct the description; without a meaningful SVD baseline, the claim in Section 4.3 that DEMUD outperforms SVD is unsupported.
  2. [4.4.1 and 4.3] The explanation stage and the discovery stage use different feature layers. Section 4.3 identifies DEMUD with fc8 features as the best discovery method, but Section 4.4.1 states that all visualizations are generated from fc6 features because fc6 explanations are more interpretable. No evidence is provided that an fc6 residual corresponds to the fc8 novelty that caused an image to be selected. The user study in Section 4.5 therefore evaluates a different detector than the one that achieved the headline discovery results. Please either add analyses linking fc6 residuals to fc8 selections (e.g., comparing explanations for the same selected images across layers, or measuring how well fc6 residuals predict fc8 novelty) or explicitly delimit the claim of 'explaining novel image discovery' to the fc6-based DEMUD detector.
  3. [4.5, Figures 17-18] The user study provides only weak and partially contradictory evidence for explanation utility. Of 20 images, only 3 (Figure 17) show significantly more 'Yes (new type of object)' votes with explanations, while 3 other images (Figure 18) show significantly fewer such votes with explanations. The paper's explanation for the latter ('users were confused') is speculative, and the significant increase in 'Not sure' responses with explanations contradicts Hypothesis 1. These results should be reported as evidence that explanations can change judgments in both directions, rather than as a clean demonstration that explanations improve users' understanding of novelty.
minor comments (5)
  1. [3.1 vs 4.1] The input image size is stated as 227x227 in Section 3.1 and 277x277 in Section 4.1; please reconcile.
  2. [3.3] The method name 'UC DeePSim' should be 'UC DeePSiM' for consistency.
  3. [4.2.2] The SIFT comparison reports the best kSIFT among tested values, which is a form of selection on the test data; this should be acknowledged as favorable to SIFT.
  4. [4.5] The paper does not report effect sizes or confidence intervals for the significant differences; given the large number of tests (20 images), a multiple-comparison correction would be helpful.
  5. [Figure 10] The annotation in panel (a) is referenced in the text but is not fully defined; please label the axes or panels consistently.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: DEMUD-VIS combines independently specified components; the residual explanation is a defined decomposition, not a fitted prediction.

full rationale

The derivation chain is self-contained. DEMUD's reconstruction and residual are explicitly defined in Eqs. (1)-(3), so the residual is exactly the unexplained feature content whose inversion the paper visualizes; this is a decomposition of the algorithm's own novelty criterion, not a prediction smuggled in from fitted data. No parameter is fitted to the evaluation data and then renamed a prediction: the mean-shift adjustment in Eq. (4) is an acknowledged preprocessing workaround for fc6 residuals, and the paper explicitly reports that fc8 residual visualization remains an open problem. The fc6/fc8 mismatch highlighted in Section 4.4.1 is a limitation on the faithfulness of the explanations, not a circular reduction, because the discovery stage and the explanation stage are evaluated separately and the visualized residual is still computed from the same DEMUD definition. The only self-citations are to the prior DEMUD algorithm, which is re-specified in full in the text, and to the authors' own ICML workshop paper, which is used as a priority note rather than as load-bearing evidence for the derivation. The user study provides independent external assessment of explanation utility. No load-bearing step reduces to its own inputs, so no circularity can be exhibited.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on three domain assumptions: transferable CNN features, reconstruction error as novelty, and feature inversion as faithful visualization. The paper offers empirical support for each, but also documents failures, especially in residual inversion. Several free parameters (kSIFT selection, mean-shift, yellow cutoffs) are fitted or hand-chosen.

free parameters (5)
  • k (number of SVD components) = min(d, n), where d is feature dimensionality and n is dataset size
    Used for both DEMUD and SVD. With full rank, SVD reconstruction error approaches zero, making the baseline comparison degenerate.
  • kSIFT = best of {3,4,5,10,15,20} per dataset
    SIFT bag-of-words cluster count was selected by reporting the best nAUC over tested values (Section 4.2.2).
  • Mean-shift offset for fc6 residuals = r' = r + 1 * (mean(x') - mean(r))
    Ad hoc transformation to make fc6 residual visualizations meaningful (Section 3.3, Eq. 4).
  • Yellow cutoff thresholds for ImageNet-Yellow = Euclidean distance < 150 from (255,255,0); school bus class cut-off 170
    Dataset construction choices that shape the novelty detection task (Section 4.1).
  • Number of selections t for nAUC = number needed to discover all classes via random, or 300, whichever is less
    Evaluation horizon choice in Eq. 6; affects reported nAUC values.
assumptions (4)
  • domain assumption CaffeNet features trained on ImageNet capture semantic content that transfers to out-of-domain images (Mars, stonefly).
    The method uses CaffeNet activations as the representation for all domains (Section 3.1, Section 4.3.2).
  • domain assumption DEMUD's reconstruction error in feature space identifies semantically novel images.
    The method relies on reconstruction error in an SVD subspace as the novelty score (Section 3.2, Eq. 1).
  • domain assumption UC DeePSiM inversion converts feature vectors, including residuals, into human-interpretable images.
    The visualization step assumes the trained up-convolutional network produces faithful images; the paper reports failures for fc6 and fc8 residuals (Section 3.3).
  • domain assumption User study responses measure the practical utility of the explanations.
    The paper evaluates explanation quality via self-reported novelty judgments on SurveyMonkey (Section 4.5).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Visualizing Image Content to Explain Novel Image Discovery." pith.science (2026). https://pith.science/paper/KR5F2QQN

@misc{pith2026190805006,
  author       = {Pith},
  title        = {Pith review of: Visualizing Image Content to Explain Novel Image Discovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KR5F2QQN}},
  note         = {Machine review of arXiv:1908.05006}
}
read the original abstract

The initial analysis of any large data set can be divided into two phases: (1) the identification of common trends or patterns and (2) the identification of anomalies or outliers that deviate from those trends. We focus on the goal of detecting observations with novel content, which can alert us to artifacts in the data set or, potentially, the discovery of previously unknown phenomena. To aid in interpreting and diagnosing the novel aspect of these selected observations, we recommend the use of novelty detection methods that generate explanations. In the context of large image data sets, these explanations should highlight what aspect of a given image is new (color, shape, texture, content) in a human-comprehensible form. We propose DEMUD-VIS, the first method for providing visual explanations of novel image content by employing a convolutional neural network (CNN) to extract image features, a method that uses reconstruction error to detect novel content, and an up-convolutional network to convert CNN feature representations back into image space. We demonstrate this approach on diverse images from ImageNet, freshwater streams, and the surface of Mars.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 38 canonical work pages

  1. [1]

    Journal of Machine Learning Research 19:1--79

    Angelino E, Larus-Stone N, Alabi D, Seltzer M, Rudin C (2018) Certifiably optimal rule lists for categorical data. Journal of Machine Learning Research 19:1--79

    work page 2018

  2. [2]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 6541--6549

    Bau D, Zhou B, Khosla A, Oliva A, Torralba A (2017) Network dissection: Quantifying interpretability of deep visual representations. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 6541--6549

    work page 2017

  3. [3]

    In: Proceedings of the IJCAI-17 Workshop on Explainable AI, pp 8--13

    Biran O, Cotton C (2017) Explanation and justification in machine learning: A survey. In: Proceedings of the IJCAI-17 Workshop on Explainable AI, pp 8--13

    work page 2017

  4. [4]

    Dr Dobb's Journal of Software Tools

    Bradski G (2000) The OpenCV Library . Dr Dobb's Journal of Software Tools

    work page 2000

  5. [5]

    ACM Computing Surveys 41(3):Article No

    Chandola V, Banerjee A, Kumar V (2009) Anomaly detection: A survey. ACM Computing Surveys 41(3):Article No. 15, doi:10.1145/1541880.1541882

    arXiv 2009

  6. [6]

    In: Proceedings of the 31st Conference on Neural Information Processing Systems, pp 6970--6979

    Dabkowski P, Gal Y (2017) Real time image saliency for black box classifiers. In: Proceedings of the 31st Conference on Neural Information Processing Systems, pp 6970--6979

    work page 2017

  7. [7]

    In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp 886--893

    Dalal N, Triggs B (2005) Histograms of oriented gradients for human detection. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp 886--893

    work page 2005

  8. [8]

    In: 2014 IEEE 30th International Conference on Data Engineering, pp 88--99, doi:10.1109/ICDE.2014.6816642

    Dang XH, Assent I, Ng RT, Zimek A, Schubert E (2014) Discriminative features for identifying and interpreting outliers. In: 2014 IEEE 30th International Conference on Data Engineering, pp 88--99, doi:10.1109/ICDE.2014.6816642

    arXiv 2014

Show all 50 references
  1. [9]

    In: Xing EP, Jebara T (eds) Proceedings of the 31st International Conference on Machine Learning, PMLR, Bejing, China, pp 647--655

    Donahue J, Jia Y, Vinyals O, Hoffman J, Zhang N, Tzeng E, Darrell T (2014) DeCAF : A deep convolutional activation feature for generic visual recognition. In: Xing EP, Jebara T (eds) Proceedings of the 31st International Conference on Machine Learning, PMLR, Bejing, China, pp 647--655

    2014

  2. [10]

    In: Lee DD, Sugiyama M, Luxburg UV, Guyon I, Garnett R (eds) Advances in Neural Information Processing Systems 29, Curran Associates, Inc., pp 658--666

    Dosovitskiy A, Brox T (2016 a ) Generating images with perceptual similarity metrics based on deep networks. In: Lee DD, Sugiyama M, Luxburg UV, Guyon I, Garnett R (eds) Advances in Neural Information Processing Systems 29, Curran Associates, Inc., pp 658--666

    2016

  3. [11]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 4829--4837

    Dosovitskiy A, Brox T (2016 b ) Inverting visual representations with convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 4829--4837

    2016

  4. [12]

    Data Mining and Knowledge Discovery 29(5):1116--1151, doi:10.1007/s10618-014-0398-2

    Duan L, Tang G, Pei J, Bailey J, Campbell A, Tang C (2015) Mining outlying aspects on numeric data. Data Mining and Knowledge Discovery 29(5):1116--1151, doi:10.1007/s10618-014-0398-2

    2015 doi

  5. [13]

    AI Magazine 4(2):44--58

    Gunning D, Aha DW (2019) DARPA 's explainable artificial intelligence program. AI Magazine 4(2):44--58

    2019

  6. [14]

    Machine Learning 3:1157--1182

    Guyon I, Elisseeff A (2003) An introduction to variable and feature selection. Machine Learning 3:1157--1182

    2003

  7. [15]

    arXiv preprint arXiv:14085093

    Jia Y, Shelhamer E, Donahue J, Karayev S, Long J, Girshick R, Guadarrama S, Darrell T (2014) Caffe: Convolutional architecture for fast feature embedding. arXiv preprint arXiv:14085093

    2014

  8. [16]

    In: Proceedings of the Innovative Applications in Artificial Intelligence Conference, pp 9484--9491

    Kerner HR, Wellington DF, Wagstaff KL, Bell JF, Amor HB (2019) Novelty detection for multispectral images with application to planetary exploration. In: Proceedings of the Innovative Applications in Artificial Intelligence Conference, pp 9484--9491

    2019

  9. [17]

    In: Lee DD, Sugiyama M, Luxburg UV, Guyon I, Garnett R (eds) Advances in Neural Information Processing Systems 29, pp 2280--2288

    Kim B, Khanna R, Koyejo O (2016) Examples are not enough, learn to criticize! C riticism for interpretability. In: Lee DD, Sugiyama M, Luxburg UV, Guyon I, Garnett R (eds) Advances in Neural Information Processing Systems 29, pp 2280--2288

    2016

  10. [18]

    In: Advances in Neural Information Processing Systems 25, pp 1097--1105

    Krizhevsky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems 25, pp 1097--1105

    2012

  11. [19]

    University of Chicago Press

    Kuhn TS (1962) The Structure of Scientific Revolutions. University of Chicago Press

    1962

  12. [20]

    In: Advances in Neural Information Processing Systems 17, MIT Press, pp 793--800

    Lim J, Ross D, Lin RS, Yang MH (2005) Incremental learning for visual tracking. In: Advances in Neural Information Processing Systems 17, MIT Press, pp 793--800

    2005

  13. [21]

    International Journal of Computer Vision 60(2):91--110, doi:10.1023/B:VISI.0000029664.99615.94

    Lowe DG (2004) Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision 60(2):91--110, doi:10.1023/B:VISI.0000029664.99615.94

    2004

  14. [22]

    Journal of the North American Benthological Society 29(3):867--874

    Lytle DA, Mart\' i nez-Mu\ n oz G, Zhang W, Larios N, Shapiro L, Paasch R, Moldenke A, Mortensen EA, Todorovic S, Dietterich TG (2010) Automated processing and identification of benthic invertebrate samples. Journal of the North American Benthological Society 29(3):867--874

    2010

  15. [23]

    Data Mining and Knowledge Discovery 32:1444--1480, doi:10.1007/s10618-018-0585-7

    Macha M, Akoglu L (2018) Explaining anomalies in groups with characterizing subspace rules. Data Mining and Knowledge Discovery 32:1444--1480, doi:10.1007/s10618-018-0585-7

    2018 doi

  16. [24]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 5188--5196

    Mahendran A, Vedaldi A (2015) Understanding deep image representations by inverting them. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 5188--5196

    2015

  17. [25]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 549--556

    Mart\' i nez-Mu\ n oz G, Zhang W, Payet N, Todorovic S, Larios N, Yamamuro A, Lytle D, Moldenke A, Mortensen E, Paasch R, Shapiro L, Dietterich T (2009) Dictionary-free categorization of very similar objects via stacked evidence trees. In: Proceedings of the IEEE Conference on...

    2009

  18. [26]

    In: 4th International Conference on Learning Representations

    Mathieu M, Couprie C, LeCun Y (2016) Deep multi-scale video prediction beyond mean square error. In: 4th International Conference on Learning Representations

    2016

  19. [27]

    In: Proceedings of the IEEE 13th International Conference on Data Mining, pp 518--527, doi:10.1109/ICDM.2013.132

    Micenkov\' a B, Ng RT, Dang XH, Assent I (2013) Explaining outliers by subspace separability. In: Proceedings of the IEEE 13th International Conference on Data Mining, pp 518--527, doi:10.1109/ICDM.2013.132

    2013 doi

  20. [28]

    Digital Signal Processing 73:1--15, doi:10.1016/j.dsp.2017.10.011

    Montavon G, Samek W, M\" u ller KR (2018) Methods for interpreting and understanding deep neural networks. Digital Signal Processing 73:1--15, doi:10.1016/j.dsp.2017.10.011

    2018 doi

  21. [29]

    In: Proceedings of the 27th International Conference on Machine Learning, pp 807--814

    Nair V, Hinton GE (2010) Rectified linear units improve restricted boltzmann machines. In: Proceedings of the 27th International Conference on Machine Learning, pp 807--814

    2010

  22. [30]

    a inen M, M\

    Ojala T, Pietik\" a inen M, M\" a enp\" a \" a T (2002) Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence 24(7):971--987

    2002

  23. [31]

    Distill doi:10.23915/distill.00007, https://distill.pub/2017/feature-visualization

    Olah C, Mordvintsev A, Schubert L (2017) Feature visualization. Distill doi:10.23915/distill.00007, https://distill.pub/2017/feature-visualization

    2017 doi

  24. [32]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 1717--1724

    Oquab M, Bottou L, Laptev I, Sivic J (2014) Learning and transferring mid-level image representations using convolutional neural networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 1717--1724

    2014

  25. [33]

    Signal Processing 99:215--249, doi:10.1016/j.sigpro.2013.12.026

    Pimentel MAF, Clifton DA, Clifton L, Tarassenko L (2014) A review of novelty detection. Signal Processing 99:215--249, doi:10.1016/j.sigpro.2013.12.026

    2014 doi

  26. [34]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp 806--813

    Razavian AS, Azizpour H, Sullivan J, Carlsson S (2014) CNN features off-the-shelf: An astounding baseline for recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp 806--813

    2014

  27. [35]

    In: Proceedings of the 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, pp 1135--1144

    Ribeiro MT, Singh S, Guestrin C (2016) `` W hy should I trust you?'' E xplaining the predictions of any classifier. In: Proceedings of the 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, pp 1135--1144

    2016

  28. [36]

    In: Proceedings of the NIPS Workshop on Critiquing and Correcting Trends in Machine Learning

    Rudin C (2018) Please stop explaining black box models for high-stakes decisions. In: Proceedings of the NIPS Workshop on Critiquing and Correcting Trends in Machine Learning

    2018

  29. [37]

    International Journal of Computer Vision 115(3):211--252

    Russakovsky O, Deng J, Su H, Krause J, Satheesh S, Ma S, Huang Z, Karpathy A, Khosla A, Bernstein M, Berg AC, Fei-Fei L (2015) ImageNet Large Scale Visual Recognition Challenge . International Journal of Computer Vision 115(3):211--252

    2015

  30. [38]

    ACM Transactions on Knowledge Discovery from Data 13(1), doi:10.1145/3230666

    Siddiqui MA, Fern A, Dietterich TG, Wong WK (2019) Sequential feature explanations for anomaly detection. ACM Transactions on Knowledge Discovery from Data 13(1), doi:10.1145/3230666

    2019 doi

  31. [39]

    Data set on Zenodo, doi:10.5281/zenodo.1049137

    Stanboli A, Wagstaff K (2017) Mars surface image ( C uriosity rover) labeled data set (version 1.0.0). Data set on Zenodo, doi:10.5281/zenodo.1049137

    2017 doi

  32. [40]

    In: Proceedings of the Workshop on Human Interpretability in Machine Learning (WHI), pp 107--113

    Wagstaff KL, Lee J (2018) Interpretable discovery in large image data sets. In: Proceedings of the Workshop on Human Interpretability in Machine Learning (WHI), pp 107--113

    2018

  33. [41]

    In: Proceedings of the Twenty-Seventh Conference on Artificial Intelligence, pp 905--911

    Wagstaff KL, Lanza NL, Thompson DR, Dietterich TG, Gilmore MS (2013) Guiding scientific discovery with explanations using DEMUD . In: Proceedings of the Twenty-Seventh Conference on Artificial Intelligence, pp 905--911

    2013

  34. [42]

    In: Proceedings of the Innovative Applications in Artificial Intelligence Conference, pp 7867--7872

    Wagstaff KL, Lu Y, Stanboli A, Grimes K, Gowda T, Padams J (2018) Deep M ars: CNN classification of M ars imagery for the PDS Imaging Atlas . In: Proceedings of the Innovative Applications in Artificial Intelligence Conference, pp 7867--7872

    2018

  35. [43]

    In: Proceedings of the ICML Deep Learning Workshop

    Yosinski J, Clune J, Nguyen A, Fuchs T, Lipson H (2015) Understanding neural networks through deep visualization. In: Proceedings of the ICML Deep Learning Workshop

    2015

  36. [44]

    In: Proceedings of the European Conference on Computer Vision, pp 818--833

    Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. In: Proceedings of the European Conference on Computer Vision, pp 818--833

    2014

  37. [45]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mi...

  38. [46]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

  39. [47]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION in...

  40. [48]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or...

  41. [49]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION in...

  42. [50]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.