Self-reinforcing Unsupervised Matching

Changshui Zhang; Jiang Lu; Lei Li

REVIEW 4 major objections 6 minor 36 references

Self-reinforcing Unsupervised Matching

T0 review · 4 major / 6 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 A self-reinforcing matching loop can annotate images in an emerging modality using only one template per class in a seen modality.

desk verdict A genuinely new DPW + self-reinforcing matching framework with strong Chinese character results, but the evaluation is partly in-sample and the seed-pair robustness gap is real; it deserves referees, not a desk rejection. read the letter →

arxiv 1909.04138 v1 pith:2S2GEKUL submitted 2019-08-23 cs.CV

Jiang Lu , Lei Li , Changshui Zhang This is my paper · ORCID

classification cs.CV

keywords cross-modalitymatchingself-reinforcinglearningdynamicpositionwarpinglocalfeatureadapterunsupervisedannotationChinesecharactertrafficsigncontinual

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 a framework, SUM, that can label images in a new, unseen visual style using only one labeled template per class from an already known style. It works for objects whose 2D layout is preserved across styles, such as Chinese characters and traffic signs. The method aligns feature matrices with an order-preserving dynamic-programming path, adapts local features across styles, and repeats matching and adaptation in a self-reinforcing loop until all images are matched. The paper reports high top-1 and top-5 match accuracies across seven Chinese-character font directions and on 42 traffic-sign classes, far above CNN, domain-adaptation, and KNN baselines.

What carries the argument

The central object is the DPW path: a two-level hierarchical warping path whose first-level nodes align rows between two feature matrices and whose second-level nodes align columns within matched rows, under boundary, monotonicity, and step-size conditions. DPW distance is computed by dynamic programming in O(HsWsHeWe) and serves as both the whole-image similarity metric for choosing matches and the source of element-level correspondences for training the local feature adapter. The second mechanism is the LoFA, a shared multi-layer perceptron that re-encodes each feature element without changing its position, optimized by the self-reinforcing SLoMa loop inside each SWIM iteration.

What would settle it

Apply SUM to a cross-modality pair where the emerging image is a 90-degree rotation of the seen template; since DPW requires order-preserving row and column alignment, the claimed matching accuracy should collapse to near chance, showing that the structure-preserving condition is load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that a fully unsupervised emerging modality can be annotated by cross-modality matching to a seen modality with one template per class, provided the objects are 2D structure-preserving. The paper proposes dynamic position warping (DPW), a hierarchical dynamic time warping for 2D feature matrices, and uses it both to compare whole images and to define local element correspondences. A learnable local feature adapter (LoFA) transfers feature elements from the emerging to the seen modality; the SLoMa algorithm alternates between finding optimal warping paths and retraining the adapter, and the SWIM algorithm progressively absorbs the most confident matches, growing from a small subset to the full set. The authors reinterpret SLoMa as a hard-expectation-maximization procedure and argue that SUM provides a route toward continual learning by reducing manual labeling effort.

Load-bearing premise

The framework depends on the emerging images being roughly 2D structure-preserving versions of the seen templates, so that an order-preserving row-and-column alignment exists, and on the first few pairs chosen by the initial identity adapter being correct enough that the self-reinforcing loop never entrenches a wrong match.

Editorial extensions

If this is right

An agent could automatically label a newly encountered font or sign style with no manual annotation, as long as it has one clean template per class in a known style.
The same framework should apply to any visual object class whose images are 2D layout-preserving, since DPW does not depend on character- or sign-specific structure.
Match accuracy starts low (roughly 3–20%) and climbs steadily as SWIM and SLoMa iterations proceed, showing that the self-reinforcing loop is what drives the gain.
Larger exploration steps in SWIM converge faster but lower final accuracy; the one-by-one exploration mode gives the steadiest, most accurate matching.
When SUM does fail, it tends to confuse characters that share local stroke units or geometric structures, indicating that matching is driven by local structure rather than global appearance.

Reading between the lines

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

A natural, paper-explicit extension is to multiple emerging modalities at once; since LoFA is trained per cross-modality direction, a multi-modality version would need shared or paired adapters, which the paper leaves for future work.
The very first SWIM iteration uses the identity LoFA, so the first absorbed pairs are chosen by raw feature distance; if one of those early pairs is wrong, the self-reinforcing loop may entrench the error, and a testable safeguard would be to compare convergence after starting from several candidate first pairs.
Because DPW is a two-dimensional time warping, the method may transfer to other aligned 2D signals, such as medical image slices or spectrograms, whenever monotonic row and column structure holds.
The feature encoder is trained only on the seen modality, so the encoder's invariance properties may matter as much as the matching machinery; ablating different encoders could isolate where the cross-modality gain actually comes from.

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.

Desk Editor's Note

read the letter

Jiang, Lei, and Zhang propose SUM, a framework for matching images across an emerging modality with only one template per class in a seen modality. The new piece is DPW, a hierarchical 2D extension of DTW, combined with a self-reinforcing loop (SWIM outer, SLoMa inner) that progressively adapts a local feature adapter. The combination is genuinely new relative to the cited DTW and domain adaptation literature, and the DPW definition and optimality proof in the appendix are solid. The toy example helps.

The Chinese character experiments are the main evidence: 7 directions, 100 random 100-way tasks, averaged top-1/top-5 accuracies. Gains over CNN, DA, and KNN baselines are large, and the error patterns align with known Chinese character confusion categories. That is a credible demonstration that the method captures local structure.

The soft spots are real. First, the final matching accuracy is computed on the same emerging feature matrices used to fit LoFA inside SWIM. Eq. 32 evaluates on the adapted substitutes of the same E_i that produced LoFA, so the result is in-sample. There is no held-out set of emerging images or per-class leave-one-out evaluation. The traffic sign result is a single run on 42 classes selected for similar backgrounds, and the Speed (80) Limit Cancel class is excluded for reasons given in the appendix. Fine, but that makes the 35/42 number much less general.

Second, the stress-test concern about the first absorbed pair is on target. Algorithm 3 seeds with an identity LoFA and absorbs the closest pair by DPW distance under the unadapted CNN encoder. If that first pair is wrong, SLoMa trains LoFA on a spurious correspondence and the loop has no verification or recovery mechanism. The EM re-derivation in Sec. 2.6.3 drops the L2 term with an unproven assertion that shortening distance along one HiPa congenitally increases distances along others. That may hold in practice, but it is not argued rigorously, and it is exactly what would need to hold for a bad seed not to be reinforced. The paper does not test what happens with an adversarial or even randomly chosen seed.

The writing is mostly clear for a methods paper, though there are typos and no code or data is released, limiting reproducibility. Who is this for? Anyone working on cross-modal matching, low-shot domain adaptation, or continual learning with emerging visual modalities. The DPW idea alone is worth citing. But as submitted, the empirical claims are stronger than the evaluation supports.

Recommendation: send it to referees. The framework deserves a careful look, and a good referee can push for held-out evaluation, seed-robustness analysis, and code release. The core idea is sound enough to be worth that effort.

Referee Report

4 major / 6 minor

Summary. The paper proposes Self-reinforcing Unsupervised Matching (SUM), a framework for cross-modality image matching under a difficult setting: one seen modality with one template per object class, and a fully unsupervised emerging modality. The method consists of a CNN feature encoder, a dynamic position warping (DPW) distance that generalizes dynamic time warping to 2D feature matrices, a local feature adapter (LoFA) implemented as an MLP, and two nested self-reinforcing loops: SLoMa (inner) optimizes LoFA by alternating DPW-based matching with feature-element regression, and SWIM (outer) progressively absorbs the most confident cross-modality pairs and retrains LoFA. The paper reports high top-1 and top-5 match accuracies across seven Chinese-character font directions and on a traffic-sign experiment, substantially outperforming CNN, domain adaptation, and KNN baselines.

Significance. If the approach is robust, it addresses a genuinely useful problem: autonomous annotation of an emerging visual modality with extremely scarce supervision. The paper introduces a sensible 2D extension of DTW with a correctness proof for the backtracking procedure (Appendix A), and the Chinese-character evaluation is extensive: 100 random trials, seven cross-font directions, and self-reinforcement trajectories analyzed in Fig. 7. However, the central self-reinforcing mechanism is not yet theoretically or empirically guaranteed: the outer loop has no protection against an incorrect first correspondence, the EM-style derivation discards a term without proof, and the traffic-sign evidence is a single curated experiment. These gaps currently prevent the paper from fully supporting its broad claims.

major comments (4)

[§2.7 (Algorithm 3, lines 3–8)] The self-reinforcing loop has no mechanism to detect or recover from an incorrect first selection: under the initial identity LoFA, the pair with the smallest DPW distance is absorbed, and if that pair is wrong, SLoMa trains LoFA on a spurious element-wise correspondence that then biases all later selections. The convergence criterion in Algorithm 2, ||w(t) - w(t-1)|| <= epsilon, only measures weight change and cannot detect a wrong seed. The paper provides no robustness analysis or experiment for this failure mode, so the central claim that SUM can autonomously annotate an emerging modality is not yet supported for the worst case. Please add either a theoretical condition under which the first selected pair is correct, an explicit verification/recovery mechanism, or experiments with deliberately incorrect or non-informative initial selections.
[§2.6.3 (Eq. 21 and Eq. 22)] The Hard-EM simplification drops the L2 term in Eq. 21 and justifies this by asserting that shortening the distance along one HiPa 'congenitally increases' distances along other HiPas. This assertion is not proven and is not generally true: if LoFA moves emerging feature elements toward the corresponding seen elements, many alternative path distances can also decrease. Since this step is the only theoretical justification for the SLoMa objective in Eq. 22, please either provide a proof or a precise condition under which the L2 term can be neglected, or reframe the derivation as a heuristic rather than an EM approximation.
[§3.2 and Appendix B] The traffic-sign experiment is a single run on 42 classes, excludes the 'Speed (80) Limit Cancel' class, and selects in-the-field images that deliberately populate a similar background with branches and sky. The reported top-1 accuracy of 35/42 and top-5 accuracy of 40/42 are therefore a controlled case study rather than a general evaluation of traffic-sign matching. The abstract and conclusion should either be revised to state this limitation explicitly or be supported by additional experiments across different sites, backgrounds, and the excluded class.
[§2.7 (Algorithm 3, lines 6–8)] Algorithm 3 never removes a selected seen template from the candidate set, so the same seen template can be matched to multiple emerging images. In the current experiments, where each class has exactly one emerging image, this permits duplicate assignments and may distort the reported match accuracy. Please enforce a bijective matching in the final decision rule, or analyze whether duplicate assignments occur in practice and what effect they have on the reported accuracy.

minor comments (6)

[§2, first paragraph] The word 'Fisrt' should be 'First'.
[§2.4 (Eq. 8)] The notation DTW(S(h_s), E(h_e)) is used before being defined; please give an explicit definition of the DTW distance between two row feature sequences.
[Appendix A] In the proof after Eq. 24, the symbol D(a,b) is used before its definition as the hierarchical accumulated distance matrix; please define the matrix D before this use.
[§2.7 (Algorithm 3, lines 7–8)] Line 7 already determines both E_l^n and S_k^n via the double arg min; line 8 recomputes S_k^n for the already chosen E_l^n and is redundant, which complicates reading.
[§3.3 and Appendix] The match accuracies reported in Fig. 5 and Fig. 7 are computed by Eq. 32 on the same emerging feature matrices used to fit the LoFA in Eq. 22. If the claim is that SUM labels the given batch, this in-sample evaluation is appropriate; if generalization to new data in the emerging modality is intended, a held-out evaluation is needed. Please state the intended claim explicitly.
[Appendix B] The description says only one template per class is available, but training uses 100 augmented copies per class for the CNN and the DA baseline; please clarify that the one-template constraint applies to the original images and that augmentation is used for training.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the SWIM/SLoMa loop is a stated self-reinforcing bootstrap with real external correctness risks, but its equations do not reduce to their inputs by construction.

full rationale

We find no circular step that meets the required standard of exhibiting a specific reduction by construction. DPW is defined independently through dynamic programming (Eqs. 6-8 and Alg. 1) with a proof of optimality in Appendix A, and it does not presuppose the matching result. The LoFA objective in Eq. 22 is the algorithm's own training loss, obtained from an EM-style Hard-EM approximation that explicitly drops the L2 term in Eq. 21; the paper gives an unproven monotonicity assertion to justify this drop. That is a correctness and robustness concern, not a circularity: the objective is not identical to the target matching result by definition, and the method could fail if the assertion is false. Similarly, the SWIM bootstrap starts with an identity LoFA and absorbs pairs under the current adapter; this is a self-training loop, and the paper explicitly calls the correspondences 'subjectively conjectural' rather than claiming they are derived from ground truth. The initial-pair correctness assumption is an uncontrolled premise and a legitimate external weakness, but it is not a case of fitted input being renamed as prediction. Evaluation via Eq. 32 uses the same emerging feature matrices that drove adaptation, but the task is transductive: the goal is to produce correspondences for exactly this set, so reporting matching accuracy on this set is the task objective rather than a held-out generalization claim. No load-bearing self-citations or imported uniqueness theorems appear; the paper's cited prior work is not used to force its central design choice. Overall, the derivation chain is self-contained as an algorithmic proposal, with its main limitations located in unsupported assumptions and bootstrap robustness rather than in circular reasoning.

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

The central claim rests on the DPW recurrence and the self-reinforcing loop. The DPW recurrence uses standard dynamic programming optimality. The loop adds hand-chosen hyperparameters (alpha, epsilon, MLP width, augmentation count) and assumes the 2D structure-preserving property plus feature transferability. The EM reinterpretation in Sec. 2.6.3 introduces a softmax model with an unspecified scale alpha and then drops the L2 term, so the optimized objective (Eq. 22) is an ad hoc simplification rather than a derived EM objective. No physical entities are invented.

free parameters (4)

exploration step size alpha = 1 (default; also tested 2, 4, 10)
Algorithm 3 adds alpha new matches per SWIM iteration; Fig. 7(c) shows larger alpha reduces final accuracy, so alpha=1 is chosen.
SLoMa/SWIM convergence threshold epsilon = 1e-3
Stopping criterion in Algorithms 2 and 3; set by hand.
LoFA hidden layer width = 400 (Chinese), 200 (traffic signs)
2-layer MLP with sigmoid and dropout; chosen without ablation.
Seen-modality augmentation size = 100 samples per class
Each template transformed by translation, shearing, and scaling to generate training data for the feature encoder; the amount is a free choice.

assumptions (5)

standard math Dynamic programming optimality of DTW/DPW recurrences
Appendix A proves optimality of the DPW path under the recurrence; relies on standard DP arguments.
domain assumption 2D structure-preserving property holds for the target objects
Definition 1 imposes hierarchical monotonicity; the paper assumes Chinese characters and traffic signs satisfy it, without a general characterization.
domain assumption CNN features trained on augmented seen templates are informative enough for cross-modality alignment
Section 2.3 freezes the first three conv blocks; no analysis of feature transferability.
ad hoc to paper Softmax distribution over HiPas with arbitrary scale alpha
Eq. 14 postulates Pr(pi) proportional to exp(-alpha D_pi); alpha is unspecified and the L2 term is later dropped, so the EM interpretation is heuristic.
domain assumption LoFA, a shared element-wise MLP, can absorb the local representation discrepancy
Section 2.5 assumes a per-element re-encoding suffices; position and context are ignored.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Self-reinforcing Unsupervised Matching." pith.science (2026). https://pith.science/paper/2S2GEKUL

@misc{pith2026190904138,
  author       = {Pith},
  title        = {Pith review of: Self-reinforcing Unsupervised Matching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2S2GEKUL}},
  note         = {Machine review of arXiv:1909.04138}
}

read the original abstract

Remarkable gains in deep learning usually rely on tremendous supervised data. Ensuring the modality diversity for one object in training set is critical for the generalization of cutting-edge deep models, but it burdens human with heavy manual labor on data collection and annotation. In addition, some rare or unexpected modalities are new for the current model, causing reduced performance under such emerging modalities. Inspired by the achievements in speech recognition, psychology and behavioristics, we present a practical solution, self-reinforcing unsupervised matching (SUM), to annotate the images with 2D structure-preserving property in an emerging modality by cross-modality matching. This approach requires no any supervision in emerging modality and only one template in seen modality, providing a possible route towards continual learning.

Figures

Figures reproduced from arXiv: 1909.04138 by the authors.

**Figure 2.** A toy matching example for DPW. The gray values are [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Diagram for (a) “optimize” and (b) “adapt” in SLoMa. In (a), the strips of [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Overall framework of SUM. SLoMa (red whirl arrow) operates inside each SWIM iteration (black whirl arrow). The subsets {Ski } nT i=1 and {Eli } nT i=1 in SWIM iteration T are formed by absorbing the potentially matched cross-modality feature matrices in {Si} N i=1 and …

**Figure 5.** Figure 5: (a) Top-1 and (b) top-5 match accuracy by different methods. Results of Chinese character matching (the front 7 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Matching results for traffic signs. For one emerging sign, its matched templates ranking from 1 to 5 are presented. The [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

**Figure 7.** Figure 7: (a) SUM performance versus SWIM iteration [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 28 canonical work pages

[1]

Deep learning,

Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015

2015
[2]

Imagenet classiﬁca- tion with deep convolutional neural networks,

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classiﬁca- tion with deep convolutional neural networks,” in NIPS, 2012, pp. 1097–1105

work page 2012
[3]

Dermatologist-level classiﬁcation of skin cancer with deep neural networks,

A. Esteva, B. Kuprel, R. A. Novoa, J. Ko, S. M. Swetter, H. M. Blau, and S. Thrun, “Dermatologist-level classiﬁcation of skin cancer with deep neural networks,” Nature, vol. 542, no. 7639, pp. 115–118, 2017

work page 2017
[4]

Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states,

T. Gebru, J. Krause, Y. Wang, D. Chen, J. Deng, E. L. Aiden, and L. Fei-Fei, “Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states,” Proceedings of the National Academy of Sciences , vol. 114, no. 50, pp. 13 108–13 113, 2017

work page 2017
[5]

An explainable deep machine vision framework for plant stress phenotyping,

S. Ghosal, D. Blystone, A. K. Singh, B. Ganapathysubramanian, A. Singh, and S. Sarkar, “An explainable deep machine vision framework for plant stress phenotyping,” Proceedings of the National Academy of Sciences, vol. 115, no. 18, pp. 4613–4618, 2018

work page 2018
[6]

Automatically identifying, count- ing, and describing wild animals in camera-trap images with deep learning,

M. S. Norouzzadeh, A. Nguyen, M. Kosmala, A. Swanson, M. S. Palmer, C. Packer, and J. Clune, “Automatically identifying, count- ing, and describing wild animals in camera-trap images with deep learning,” Proceedings of the National Academy of Sciences , vol. 115, no. 25, pp. E5716–E5725, 2018

work page 2018
[7]

Self-paced learning with diversity,

L. Jiang, D. Meng, S.-I. Yu, Z. Lan, S. Shan, and A. Hauptmann, “Self-paced learning with diversity,” in NIPS, 2014

work page 2014
[8]

Fast mixing markov chains for strongly rayleigh measures, dpps, and constrained sampling,

C. Li, S. Sra, and S. Jegelka, “Fast mixing markov chains for strongly rayleigh measures, dpps, and constrained sampling,” in NIPS, 2016

work page 2016

Show all 36 references

[9]

Low-shot learning from imaginary data,

Y.-X. Wang, R. Girshick, M. Hebert, and B. Hariharan, “Low-shot learning from imaginary data,” in CVPR, 2018, pp. 7278–7286

2018
[10]

Autoaugment: Learning augmentation policies from data,

E. D. Cubuk, B. Zoph, D. Mane, V . Vasudevan, and Q. V . Le, “Autoaugment: Learning augmentation policies from data,” arXiv preprint arXiv:1805.09501, 2018

2018 arXiv
[11]

Unbiased look at dataset bias,

A. Torralba and A. A. Efros, “Unbiased look at dataset bias,” in CVPR, 2011

2011
[12]

Domain adaptation for object recognition: An unsupervised approach,

R. Gopalan, R. Li, and R. Chellappa, “Domain adaptation for object recognition: An unsupervised approach,” in ICCV, 2011

2011
[13]

Lifelong machine learning,

Z. Chen and B. Liu, “Lifelong machine learning,” Synthesis Lectures on Artiﬁcial Intelligence and Machine Learning , vol. 10, no. 3, pp. 1–145, 2016

2016
[14]

Data- driven visual similarity for cross-domain image matching,

A. Shrivastava, T. Malisiewicz, A. Gupta, and A. A. Efros, “Data- driven visual similarity for cross-domain image matching,” ACM Transactions on Graphics (ToG), vol. 30, no. 6, pp. 154(1)–154(9), 2011

2011
[15]

Distance metric learning with application to clustering with side-information,

E. P . Xing, M. I. Jordan, S. J. Russell, and A. Y. Ng, “Distance metric learning with application to clustering with side-information,” in NIPS, 2003

2003
[16]

Distance metric learning for large margin nearest neighbor classiﬁcation,

K. Q. Weinberger and L. K. Saul, “Distance metric learning for large margin nearest neighbor classiﬁcation,” Journal of Machine Learning Research, vol. 10, no. Feb, pp. 207–244, 2009

2009
[17]

Writer adaptation with style transfer mapping,

X.-Y. Zhang and C.-L. Liu, “Writer adaptation with style transfer mapping,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 7, pp. 1773–1787, 2013

2013
[18]

Deep domain confusion: Maximizing for domain invariance,

E. Tzeng, J. Hoffman, N. Zhang, K. Saenko, and T. Darrell, “Deep domain confusion: Maximizing for domain invariance,” arXiv preprint arXiv:1412.3474, 2014

2014 arXiv
[19]

Dynamic programming algorithm op- timization for spoken word recognition,

H. Sakoe and S. Chiba, “Dynamic programming algorithm op- timization for spoken word recognition,” IEEE Transactions on Acoustics, Speech, and Signal Processing , vol. 26, no. 1, pp. 43–49, 1978

1978
[20]

Wundt, Lectures on human and animal psychology (Psychology revivals)

W. Wundt, Lectures on human and animal psychology (Psychology revivals). Routledge, 1912

1912
[21]

Stimulus similarity as a determinant of pavlovian conditioning,

R. A. Rescorla and D. R. Furrow, “Stimulus similarity as a determinant of pavlovian conditioning,” Journal of Experimental Psychology: Animal Behavior Processes , vol. 3, no. 3, pp. 203–215, 1977

1977
[22]

Behavior theory and the models of man

A. Bandura, “Behavior theory and the models of man.” American Psychologist, vol. 29, no. 12, pp. 859–869, 1974

1974
[23]

Happiness runs in a circular motion: Evidence for a positive feedback loop between prosocial spending and happiness,

L. B. Aknin, E. W. Dunn, and M. I. Norton, “Happiness runs in a circular motion: Evidence for a positive feedback loop between prosocial spending and happiness,” Journal of Happiness Studies , vol. 13, no. 2, pp. 347–355, 2012

2012
[24]

J. R. Anderson, Cognitive psychology and its implications. Macmillan, 2005

2005
[25]

Very deep convolutional networks for large-scale image recognition,

K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2014

2014 arXiv
[26]

Rectiﬁed linear units improve restricted boltzmann machines,

V . Nair and G. E. Hinton, “Rectiﬁed linear units improve restricted boltzmann machines,” in ICML, 2010

2010
[27]

C. M. Bishop et al., Neural networks for pattern recognition. Oxford university press, 1995

1995
[28]

E. L. Lehmann and G. Casella, Theory of point estimation. Springer Science & Business Media, 2006

2006
[29]

The expectation-maximization algorithm,

T. K. Moon, “The expectation-maximization algorithm,”IEEE Signal processing magazine, vol. 13, no. 6, pp. 47–60, 1996

1996
[30]

A gentle tutorial of the em algorithm and its application to parameter estimation for gaussian mixture and hidden markov models,

J. A. Bilmes et al. , “A gentle tutorial of the em algorithm and its application to parameter estimation for gaussian mixture and hidden markov models,” International Computer Science Institute , vol. 4, no. 510, p. 126, 1998

1998
[31]

Semantic radicals contribute to the visual identiﬁcation of chinese characters,

L. B. Feldman and W. W. Siok, “Semantic radicals contribute to the visual identiﬁcation of chinese characters,” Journal of Memory and Language, vol. 40, no. 4, pp. 559–576, 1999

1999
[32]

The german trafﬁc sign recognition benchmark: a multi-class classiﬁcation competition,

J. Stallkamp, M. Schlipsing, J. Salmen, and C. Igel, “The german trafﬁc sign recognition benchmark: a multi-class classiﬁcation competition,” in IJCNN, 2011

2011
[33]

Incorporating nesterov momentum into adam,

T. Dozat, “Incorporating nesterov momentum into adam,” in ICLR- Workshop, 2016

2016
[34]

Adam: A method for stochastic optimiza- tion,

D. P . Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,” in ICLR, 2015

2015
[35]

Long short-term memory,

S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

1997
[36]

Reducing the dimension- ality of data with neural networks,

G. E. Hinton and R. R. Salakhutdinov, “Reducing the dimension- ality of data with neural networks,” science, vol. 313, no. 5786, pp. 504–507, 2006. Jiang Lu received the B.S. degree from Tsinghua University, Beijing, China, in 2013, where he is currently working toward the Ph....

2006

Pith tools

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

[1] [1]

Deep learning,

Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015

2015

[2] [2]

Imagenet classiﬁca- tion with deep convolutional neural networks,

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classiﬁca- tion with deep convolutional neural networks,” in NIPS, 2012, pp. 1097–1105

work page 2012

[3] [3]

Dermatologist-level classiﬁcation of skin cancer with deep neural networks,

A. Esteva, B. Kuprel, R. A. Novoa, J. Ko, S. M. Swetter, H. M. Blau, and S. Thrun, “Dermatologist-level classiﬁcation of skin cancer with deep neural networks,” Nature, vol. 542, no. 7639, pp. 115–118, 2017

work page 2017

[4] [4]

Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states,

T. Gebru, J. Krause, Y. Wang, D. Chen, J. Deng, E. L. Aiden, and L. Fei-Fei, “Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states,” Proceedings of the National Academy of Sciences , vol. 114, no. 50, pp. 13 108–13 113, 2017

work page 2017

[5] [5]

An explainable deep machine vision framework for plant stress phenotyping,

S. Ghosal, D. Blystone, A. K. Singh, B. Ganapathysubramanian, A. Singh, and S. Sarkar, “An explainable deep machine vision framework for plant stress phenotyping,” Proceedings of the National Academy of Sciences, vol. 115, no. 18, pp. 4613–4618, 2018

work page 2018

[6] [6]

Automatically identifying, count- ing, and describing wild animals in camera-trap images with deep learning,

M. S. Norouzzadeh, A. Nguyen, M. Kosmala, A. Swanson, M. S. Palmer, C. Packer, and J. Clune, “Automatically identifying, count- ing, and describing wild animals in camera-trap images with deep learning,” Proceedings of the National Academy of Sciences , vol. 115, no. 25, pp. E5716–E5725, 2018

work page 2018

[7] [7]

Self-paced learning with diversity,

L. Jiang, D. Meng, S.-I. Yu, Z. Lan, S. Shan, and A. Hauptmann, “Self-paced learning with diversity,” in NIPS, 2014

work page 2014

[8] [8]

Fast mixing markov chains for strongly rayleigh measures, dpps, and constrained sampling,

C. Li, S. Sra, and S. Jegelka, “Fast mixing markov chains for strongly rayleigh measures, dpps, and constrained sampling,” in NIPS, 2016

work page 2016