Pith. sign in

REVIEW 4 major objections 6 minor 48 references

A multi-layer temporal convolution network with per-code attention raises medical coding F1 by 9% and recall by 28% on common ICD labels.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-31 00:32 UTC pith:M7CVU3U7

load-bearing objection Clean 2019 engineering combo of TCN + label attention; real-looking top-50 gains, but the big recall number is under-supported and partly an operating-point story. the 4 major comments →

arxiv 2607.25129 v1 pith:M7CVU3U7 submitted 2026-07-27 cs.CL cs.LG

Deep Label-Wise Attentive Temporal Convolutional Networks Improve Medical Coding

Muhammed Yavuz Nuzumlal{\i} , Alexander Fabbri , Irene Li , Dragomir Radev This is my paper
classification cs.CL cs.LG
keywords medical codingICD classificationtemporal convolutional networkslabel-wise attentionmulti-label text classificationMIMIC-IIIclinical NLPdischarge summaries
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Medical coding asks a system to assign many diagnosis and procedure codes from a single long hospital note, where evidence for different codes can sit far apart. The authors treat this as multi-label text classification and show that a stack of temporal convolutional layers, which expand their view of the document exponentially, followed by a separate attention pass for each code, captures both global context and code-specific focus better than earlier local convolutional models. On the fifty most frequent ICD codes in MIMIC-III discharge summaries, their LATCN model improves average F1 by about nine percent and recall by about twenty-eight percent over the prior best system. They argue that the recall gain matters more in clinical decision support, because missing a true code is costlier than offering an extra candidate for a human to filter. The result positions long-range dilated convolution plus label-wise attention as a practical route to more useful automated coding aids.

Core claim

On the MIMIC-III top-50 ICD label set, a multi-layer temporal convolutional network followed by label-wise attention (LATCN) significantly outperforms the previous state-of-the-art attentional CNN model, raising average F1 by roughly 9% and micro-recall from 0.584 to 0.710. The authors treat the large recall lift as the clinically more important result for decision-support use.

What carries the argument

LATCN: residual temporal convolutional blocks with exponentially growing dilations that build a long-range document matrix, followed by a learned per-label attention matrix that produces one dense vector per ICD code for a final sigmoid prediction.

Load-bearing premise

The gains are measured only on the fifty most common codes, with notes truncated at 2,500 tokens and only discharge summaries used; full-label-set results are left as future work.

What would settle it

Train and evaluate the identical LATCN architecture on the full MIMIC-III ICD label set under the same splits; if the F1 and recall lifts over the prior best model vanish or reverse on the long-tail codes, the central performance claim does not hold for real coding workloads.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • High-recall automated coding can surface more true diagnoses and procedures for human review, lowering the chance a code is missed entirely.
  • Recall-first design becomes a defensible default for clinical NLP decision-support tools.
  • Multi-layer dilated convolutions are a viable encoder for long clinical notes in place of local CNNs or recurrent nets.
  • Label-specific attention remains useful even when the encoder already sees global context.
  • Measuring the same architecture on the full ICD label set and with hierarchical code structure is the direct next test the paper itself flags.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same TCN-plus-per-label-attention pattern may transfer to other long-document multi-label tasks outside medicine, such as multi-topic legal or scientific tagging.
  • Because precision falls while recall rises, the practical deployment value is as a high-recall first-pass filter for humans rather than a fully automatic coder.
  • Adding contextual embeddings or hierarchy-aware losses, both noted as future work, is a concrete next experiment for closing the remaining gap on rare codes.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes LATCN, a multi-label text classification model for automatic ICD-9 coding that stacks a multi-layer temporal convolutional network (with exponentially dilated residual blocks) under the label-wise attention mechanism of Mullenbach et al. (2018). The motivation is that dilated convolutions give a large receptive field suited to long discharge summaries, while per-label attention preserves label-specific document views. On the MIMIC-III top-50 label benchmark (discharge summaries truncated to 2,500 tokens), the authors report micro-F1 0.681 and micro-recall 0.710 versus 0.633 and 0.584 for the previous state of the art DR-CAML (Table 2), and argue the recall gain (~28% relative) is the clinically important improvement. The manuscript carries an explicit note that the work was done in 2019 and is posted as a record of that work rather than a current account of the field.

Significance. If the recall gain survives matched-threshold evaluation, the result is a meaningful contribution to the 2019-era medical coding literature: LATCN is a simple, parameter-modest architecture (4 residual blocks, 100 channels, k=4) whose exponentially dilated receptive field is well motivated for long discharge summaries (median 1,388 words), and the label-wise attention preserves the interpretability property of CAML. The paper's explicit framing as an archival record of 2019 work is honest and appropriately scopes its novelty claims. However, the work is evaluated only on the 50 most frequent labels with single-run results, so its evidentiary weight for current practice is limited, and it neither ships code nor provides the threshold analysis needed to certify the headline numbers.

major comments (4)
  1. [§4, Table 2] Table 2 / Eq. (4): the model outputs per-label sigmoid probabilities, but the binarization threshold used to compute the reported precision, recall, and F1 is never stated anywhere in the manuscript. This is load-bearing for the central claim: relative to DR-CAML, LATCN gains 12.6 micro-recall points (0.710 vs 0.584) while losing 3.6 micro-precision points (0.655 vs 0.691). That joint movement is exactly the signature of a less conservative operating point rather than necessarily a better ranker, and the paper's own explanation ('generalization', §4) does not rule the threshold confound out. Notably, the threshold-insensitive comparison is much narrower: micro-AUC 0.931 vs 0.916. The authors should state the threshold, and ideally report PR curves or max-F1 over thresholds for LATCN and DR-CAML under matched operating points, so the recall claim can be separated from calibration.
  2. [§3.3, §4] §3.3 / §4: all results come from a single 200-epoch training run with hyperparameters grid-searched on the development set. No seed variance, error bars, or significance tests are reported, yet the abstract and §4 describe the gains as 'significantly better' and 'remarkable'. On this benchmark, seed-to-seed variation of 1–2 micro-F1 points is routine, which would materially erode the 4.8-point micro-F1 gap over DR-CAML. At minimum the word 'significantly' should be removed or supported; ideally the model should be retrained over several seeds with mean ± std reported.
  3. [Table 2] Table 2: the CAML, DR-CAML, LEAM, CNN, Bi-GRU, and LR rows are attributed to prior papers, which strongly suggests the numbers were transcribed from Mullenbach et al. (2018) and related work rather than re-run under identical preprocessing. The manuscript does not state whether the train/dev/test splits, tokenization, and truncation (2,500 tokens, §3.1) match those used to produce the baseline numbers. Any mismatch silently inflates or deflates the comparison. The provenance of each baseline row, and the exact split used, must be stated explicitly.
  4. [§3.1, title/abstract] §3.1 and title/abstract: evaluation is restricted to the 50 most frequent ICD-9 labels, with the full-label-set evaluation explicitly deferred to future work (§6). The ~50-code setting covers only the head of the distribution, while real coding workloads are dominated by long-tail codes, and it is known from subsequent literature that relative model rankings on MIMIC-III top-50 do not always transfer to the full label set. The title ('Improve Medical Coding') and the clinical decision-support framing in §3.2 and §4 are broader than the evidence. The claims should be scoped to the top-50 setting, or full-label results added.
minor comments (6)
  1. [§2, Eq. (2)] Eq. (2): the softmax is written self-referentially (α_ℓ = exp(α_ℓ) / Σ_n exp(α_ℓn)) and contains a typesetting artifact '(sof tmax)'. Presumably the intended form is α_ℓn = exp(a_ℓn) / Σ_m exp(a_ℓm) where a_ℓn is the (n,ℓ) entry of A from Eq. (1). Please fix the notation.
  2. [Table 2] Table 2: P@5/R@5 and several precision/recall cells are left blank (dashes) for LR, CNN, Bi-GRU, and LEAM. Either fill these from the cited sources or note why they are unavailable; the partial rows make cross-model comparison harder to read.
  3. [§2] §2: the receptive field formula 2×(2^{ℓ+1}−1)×k appears to assume two convolutions per residual block and stride 1; please make the indexing convention explicit (whether ℓ starts at 0) so the formula can be checked against the 4-block architecture in Figure 1.
  4. [§3.1, §3.3] §3.1/§3.3: train/development/test split sizes and the source of the split (e.g., the standard MIMIC-III top-50 split of Mullenbach et al.) are not given; nor are the word2vec embedding dimension d_e or whether embeddings are fine-tuned. These are needed for reproduction.
  5. [§3.3] No code release or reproducibility artifact is mentioned. Given the small model and standard dataset, releasing training code and the selected configuration would substantially strengthen the paper's value as a record of the work.
  6. [§1, References, Figure 1] Minor: 'state-or-the-art' (§1) should be 'state-of-the-art'; the Schwenk et al. reference is missing a year in the citation; Figure 1 labels channel width and dilation but the figure could also indicate the 2,500-token input length for self-containment.

Circularity Check

0 steps flagged

No circularity: standard supervised multi-label evaluation on held-out MIMIC-III; metrics are not forced by construction.

full rationale

The paper's load-bearing claim is an empirical comparison (LATCN vs prior baselines on MIMIC-III top-50 ICD codes: higher micro/macro F1 and recall in Table 2). The derivation chain is ordinary supervised learning: word2vec embeddings, multi-layer TCN (Bai et al.), label-wise attention (Mullenbach et al.), BCE loss (Eq. 5), sigmoid outputs (Eq. 4), hyperparameters selected on a development set, metrics on held-out labels. Nothing in the architecture, loss, or attention equations defines F1/recall in terms of the claimed improvement, nor is any fitted scalar renamed as a prediction of a related quantity. Citations to TCN and CAML/DR-CAML are external architectural and baseline sources with no author overlap that would make the result self-justifying. Threshold choice, single-run variance, and top-50 scope are validity/robustness concerns, not circular reductions. Score 0; steps empty.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 1 invented entities

The claim rests on standard multi-label classification assumptions, the MIMIC-III discharge-summary setup used by prior work, and ordinary neural-net design choices (TCN dilations, label-wise attention, BCE). No new physical entities. Free parameters are the usual architecture and training hyperparameters selected on a development set.

free parameters (6)
  • TCN depth ℓ = 4
    Number of residual blocks chosen by grid search on dev set; controls receptive field size.
  • channel width d_ch = 100
    Convolution channel count selected by grid search.
  • filter size k = 4
    Kernel width selected by grid search.
  • dropout rate = 0.4
    Regularization strength taken from TCN paper suggestions / tuning.
  • learning rate schedule = 0.002 → 0.0001
    Adam with reduce-on-plateau from 0.002 toward 0.0001; affects final metric numbers.
  • max document length = 2500
    Texts truncated at 2,500 tokens; modeling choice that discards tail context.
axioms (5)
  • domain assumption Binary cross-entropy over independent label sigmoids is an appropriate training objective for multi-label ICD assignment.
    Section 2, Loss Function; standard but ignores label hierarchy and exclusivity constraints among ICD codes.
  • domain assumption Discharge summaries alone, truncated, contain sufficient signal for the top-50 codes.
    Section 3.1; follows prior studies but excludes other note types and long-tail codes.
  • domain assumption Exponentially dilated TCN residual blocks yield a usable global document representation for long clinical text.
    Section 2 citing Bai et al. 2018; transferred from sequence-modeling benchmarks without clinical-specific proof.
  • domain assumption Label-wise attention (Mullenbach et al. 2018) produces meaningful per-code document vectors when placed on TCN features.
    Section 2 equations (1)–(4); mechanism imported from CAML and assumed to transfer.
  • ad hoc to paper Higher recall is more valuable than higher precision for clinical decision-support coding tools.
    Stated in abstract, §3.2 and Discussion; normative preference that frames the main positive result, not an empirical measurement in the paper.
invented entities (1)
  • LATCN (Label-wise Attentive Temporal Convolutional Network) no independent evidence
    purpose: Name for the stacked TCN + label-wise attention architecture evaluated on MIMIC-III.
    Composition of existing TCN and CAML-style attention; no new mathematical object beyond the combination.

pith-pipeline@v1.2.0-grok45-kimik3 · 11576 in / 3480 out tokens · 55104 ms · 2026-07-31T00:32:52.549898+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Deep Label-Wise Attentive Temporal Convolutional Networks Improve Medical Coding." pith.science (2026). https://pith.science/paper/M7CVU3U7

@misc{pith2026260725129,
  author       = {Pith},
  title        = {Pith review of: Deep Label-Wise Attentive Temporal Convolutional Networks Improve Medical Coding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M7CVU3U7}},
  note         = {Machine review of arXiv:2607.25129}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Medical coding is the task of assigning a set of diagnosis and procedure codes for a hospitalization using recorded notes. It requires aggregating information from different parts of the text and focus to different sections for each individual code, making it a very difficult problem even for professional human coders. We model the task as a multi-label text classification problem. To overcome the mentioned difficulties, we propose a deep neural model consisting of a multi-layer temporal convolution network (TCN) followed by label-wise attention. While multi-layer TCN helps extract a global document representation with the ability to learn relations over very long sequences, label-specific attention mechanism allows the model to focus on different aspects of the same document for each individual label. Our method achieves significantly better F-1 scores (9% increase) compared to the previous state-of-the-art model, with a remarkable increase in recall score (28% increase), which we believe is the more important metric for a clinical decision support setting.

Figures

Figures reproduced from arXiv: 2607.25129 by Alexander Fabbri, Dragomir Radev, Irene Li, Muhammed Yavuz Nuzumlal{\i}.

Figure 1
Figure 1. Figure 1: High level diagram of the proposed LATCN model. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 10 linked inside Pith

  1. [1]

    Advances in Neural Information Processing Systems 26 , editor =

    Distributed Representations of Words and Phrases and their Compositionality , author =. Advances in Neural Information Processing Systems 26 , editor =. 2013 , publisher =

    2013

  2. [2]

    Manning , booktitle =

    Jeffrey Pennington and Richard Socher and Christopher D. Manning , booktitle =. GloVe: Global Vectors for Word Representation , year =

  3. [3]

    Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    Deep Contextualized Word Representations , author=. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    2018

  4. [4]

    arXiv preprint arXiv:1810.04805 , year=

    Bert: Pre-training of deep bidirectional transformers for language understanding , author=. arXiv preprint arXiv:1810.04805 , year=

    Pith/arXiv arXiv

  5. [5]

    Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    Explainable Prediction of Medical Codes from Clinical Text , author=. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    2018

  6. [6]

    arXiv preprint arXiv:1803.01271 , year=

    An empirical evaluation of generic convolutional and recurrent networks for sequence modeling , author=. arXiv preprint arXiv:1803.01271 , year=

    Pith/arXiv arXiv

  7. [7]

    npj Digital Medicine , number =

    Rajkomar, Alvin and Oren, Eyal and Chen, Kai and Dai, Andrew M and Hajaj, Nissan and Hardt, Michaela and Liu, Peter J and Liu, Xiaobing and Marcus, Jake and Sun, Mimi and Sundberg, Patrik and Yee, Hector and Zhang, Kun and Zhang, Yi and Flores, Gerardo and Duggan, Gavin E and Irvine, Jamie and Le, Quoc and Litsch, Kurt and Mossin, Alexander and Tansuwan, ...

  8. [8]

    Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    EMR Coding with Semi-Parametric Multi-Head Matching Networks , author=. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , volume=

    2018

  9. [9]

    Workshops at the Thirty-Second AAAI Conference on Artificial Intelligence , year=

    Multi-label classification of patient notes: case study on ICD code assignment , author=. Workshops at the Thirty-Second AAAI Conference on Artificial Intelligence , year=

  10. [10]

    Advances in Neural Information Processing Systems , pages=

    Distilled Wasserstein learning for word embedding and topic modeling , author=. Advances in Neural Information Processing Systems , pages=

  11. [11]

    arXiv preprint arXiv:1810.13348 , year=

    Multimodal Machine Learning for Automated ICD Coding , author=. arXiv preprint arXiv:1810.13348 , year=

    Pith/arXiv arXiv

  12. [12]

    Advances in Neural Information Processing Systems 31 , editor =

    MiME: Multilevel Medical Embedding of Electronic Health Records for Predictive Healthcare , author =. Advances in Neural Information Processing Systems 31 , editor =. 2018 , publisher =

    2018

  13. [13]

    Joint Embedding of Words and Labels for Text Classification

    Wang, Guoyin and Li, Chunyuan and Wang, Wenlin and Zhang, Yizhe and Shen, Dinghan and Zhang, Xinyuan and Henao, Ricardo and Carin, Lawrence. Joint Embedding of Words and Labels for Text Classification. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2018

    2018

  14. [14]

    Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , pages=

    Few-shot and zero-shot multi-label learning for structured label spaces , author=. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , pages=

    2018

  15. [15]

    arXiv preprint arXiv:1809.05679 , year=

    Graph Convolutional Networks for Text Classification , author=. arXiv preprint arXiv:1809.05679 , year=

    Pith/arXiv arXiv

  16. [16]

    arXiv preprint arXiv:1812.01207 , year=

    Practical Text Classification With Large Pre-Trained Language Models , author=. arXiv preprint arXiv:1812.01207 , year=

    Pith/arXiv arXiv

  17. [17]

    Scientific data , volume=

    MIMIC-III, a freely accessible critical care database , author=. Scientific data , volume=. 2016 , publisher=

    2016

  18. [18]

    arXiv preprint arXiv:1901.08746 , year=

    BioBERT: pre-trained biomedical language representation model for biomedical text mining , author=. arXiv preprint arXiv:1901.08746 , year=

    Pith/arXiv arXiv 1901

  19. [19]

    Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=

    Hierarchical attention networks for document classification , author=. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=

    2016

  20. [20]

    and Tarau, P

    Mihalcea, R. and Tarau, P. , biburl =. Proceedings of

  21. [21]

    Khapra and Anirban Laha and Balaraman Ravindran , title =

    Preksha Nema and Mitesh M. Khapra and Anirban Laha and Balaraman Ravindran , title =. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics,. 2017 , url =. doi:10.18653/v1/P17-1098 , timestamp =

    doi:10.18653/v1/p17-1098 2017

  22. [22]

    Towards the Use of Deep Reinforcement Learning with Global Policy for Query-based Extractive Summarisation , booktitle =

    Moll. Towards the Use of Deep Reinforcement Learning with Global Policy for Query-based Extractive Summarisation , booktitle =. 2017 , url =

    2017

  23. [23]

    Using Random Walks for Question-focused Sentence Retrieval , booktitle =

    Otterbacher, Jahna and Erkan, G\". Using Random Walks for Question-focused Sentence Retrieval , booktitle =. 2005 , location =. doi:10.3115/1220575.1220690 , acmid =

    arXiv 2005

  24. [24]

    Query-Based Abstractive Summarization Using Neural Networks , journal =

    Johan Hasselqvist and Niklas Helmertz and Mikael K. Query-Based Abstractive Summarization Using Neural Networks , journal =. 2017 , url =

    2017

  25. [25]

    , biburl =

    Erkan, Günes and Radev, Dragomir R. , biburl =. LexRank: Graph-based Lexical Centrality as Salience in Text Summarization. , url =. J. Artif. Intell. Res. (JAIR) , keywords =

  26. [26]

    Neural Machine Translation by Jointly Learning to Align and Translate , url =

    Bahdanau, Dzmitry and Cho, Kyunghyun and Bengio, Yoshua , biburl =. Neural Machine Translation by Jointly Learning to Align and Translate , url =. arXiv , arxivid =:1409.0473v7 , file =

    Pith/arXiv arXiv

  27. [27]

    arXiv preprint arXiv:1805.04576 , year=

    Domain Adapted Word Embeddings for Improved Sentiment Classification , author=. arXiv preprint arXiv:1805.04576 , year=

    Pith/arXiv arXiv

  28. [28]

    arXiv preprint arXiv:1801.06146 , year=

    Universal language model fine-tuning for text classification , author=. arXiv preprint arXiv:1801.06146 , year=

    Pith/arXiv arXiv

  29. [29]

    Advances in Neural Information Processing Systems , pages=

    Uncertainty-Aware Attention for Reliable Interpretation and Prediction , author=. Advances in Neural Information Processing Systems , pages=

  30. [30]

    Advances in Neural Information Processing Systems , pages=

    Streaming weak submodularity: Interpreting neural networks on the fly , author=. Advances in Neural Information Processing Systems , pages=

  31. [31]

    Advances in Neural Information Processing Systems , pages=

    Retain: An interpretable predictive model for healthcare using reverse time attention mechanism , author=. Advances in Neural Information Processing Systems , pages=

  32. [32]

    Advances in Neural Information Processing Systems , pages=

    A unified approach to interpreting model predictions , author=. Advances in Neural Information Processing Systems , pages=

  33. [33]

    arXiv preprint arXiv:1703.03130 , year=

    A structured self-attentive sentence embedding , author=. arXiv preprint arXiv:1703.03130 , year=

    Pith/arXiv arXiv

  34. [34]

    Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , pages=

    Why should i trust you?: Explaining the predictions of any classifier , author=. Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , pages=. 2016 , organization=

    2016

  35. [35]

    and Salakhutdinov, Ruslan and Manning, Christopher D

    Yang, Zhilin and Qi, Peng and Zhang, Saizheng and Bengio, Yoshua and Cohen, William W. and Salakhutdinov, Ruslan and Manning, Christopher D. , booktitle=

  36. [36]

    Advances in Neural Information Processing Systems 29 , editor =

    Examples are not enough, learn to criticize! Criticism for Interpretability , author =. Advances in Neural Information Processing Systems 29 , editor =. 2016 , publisher =

    2016

  37. [37]

    Advances in Neural Information Processing Systems , pages=

    Neural Interaction Transparency (NIT): Disentangling Learned Interactions for Improved Interpretability , author=. Advances in Neural Information Processing Systems , pages=

  38. [38]

    Proceedings of the IEEE International Conference on Computer Vision , pages=

    Grad-cam: Visual explanations from deep networks via gradient-based localization , author=. Proceedings of the IEEE International Conference on Computer Vision , pages=

  39. [39]

    Proceedings of the 27th international conference on machine learning (ICML-10) , pages=

    Rectified linear units improve restricted boltzmann machines , author=. Proceedings of the 27th international conference on machine learning (ICML-10) , pages=

  40. [40]

    2014 , archivePrefix=

    Adam: A Method for Stochastic Optimization , author=. 2014 , archivePrefix=

    2014

  41. [41]

    The Journal of Machine Learning Research , volume=

    Dropout: a simple way to prevent neural networks from overfitting , author=. The Journal of Machine Learning Research , volume=. 2014 , publisher=

    2014

  42. [42]

    Advances in Neural Information Processing Systems , pages=

    Weight normalization: A simple reparameterization to accelerate training of deep neural networks , author=. Advances in Neural Information Processing Systems , pages=

  43. [43]

    Very Deep Convolutional Networks for Text Classification , booktitle =

    Holger Schwenk and Lo. Very Deep Convolutional Networks for Text Classification , booktitle =

  44. [44]

    CoRR , volume =

    Ronghui You and Suyang Dai and Zihan Zhang and Hiroshi Mamitsuka and Shanfeng Zhu , title =. CoRR , volume =. 2018 , archivePrefix =

    2018

  45. [45]

    Deep Learning for Extreme Multi-label Text Classification , booktitle =

    Jingzhou Liu and Wei. Deep Learning for Extreme Multi-label Text Classification , booktitle =

  46. [46]

    CoRR , volume =

    Youwei Song and Jiahai Wang and Tao Jiang and Zhiyue Liu and Yanghui Rao , title =. CoRR , volume =. 2019 , archivePrefix =

    2019

  47. [47]

    CoRR , volume =

    Xiang Jiang and Mohammad Havaei and Gabriel Chartrand and Hassan Chouaib and Thomas Vincent and Andrew Jesson and Nicolas Chapados and Stan Matwin , title =. CoRR , volume =. 2018 , archivePrefix =

    2018

  48. [48]

    Hierarchical Attentional Hybrid Neural Networks for Document Classification , journal =

    Jader Abreu and Luis Fred and David Mac. Hierarchical Attentional Hybrid Neural Networks for Document Classification , journal =. 2019 , archivePrefix =

    2019