Pith. sign in

REVIEW 3 major objections 6 minor 112 references

Spatiotemporal deep learning models for detection of rapid intensification in cyclones

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read LSTM-generated synthetic cyclone tracks and wind intensities substantially improve detection of rapid intensification events, lifting F1 scores from 0.339 to 0.651 in the South Pacific and 0.487 to 0.657 in the South Indian Ocean.

desk verdict The DA-M-LSTM improvement is real in the tables, but the paper never establishes that the synthetic samples were generated outside the classifier's test period, so the headline result is unverified. read the letter →

arxiv 2506.08397 v1 pith:VN3WKTPV submitted 2025-06-10 cs.LG cs.AIstat.ML

Vamshika Sutar , Amandeep Singh , Rohitash Chandra This is my paper
classification cs.LGcs.AIstat.ML
keywords cyclonerapidintensificationclassimbalancedataaugmentationLSTMspatiotemporalsyntheticgenerationtropicaldetectiondeeplearning
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

This paper argues that the crippling class imbalance in cyclone rapid-intensification (RI) detection can be countered by generating synthetic cyclone data with a deep learning model rather than with standard oversampling techniques. The authors build an LSTM that takes the previous four six-hourly track points (latitude, longitude, wind intensity) of an RI cyclone and generates the next four, producing plausible continuations of real RI tracks. Adding these synthetic examples to the training set of a multivariate LSTM classifier lifts RI F1 scores from 0.339 to 0.651 in the South Pacific and from 0.487 to 0.657 in the South Indian Ocean. The paper also claims that spatial coordinates are critical inputs, since the track-aware multivariate model outperforms the wind-only univariate model in key comparisons. A sympathetic reader should care because rare extreme events are exactly where prediction models fail, and the framework suggests a path to synthesizing training data for other spatiotemporal extremes.

What carries the argument

The load-bearing machinery is the LSTM-based data augmentation generator. It receives four time steps (24 hours) of three features—latitude, longitude, and wind intensity—and outputs the next four time steps, so it is trained with mean-squared error on both RI and non-RI cyclone segments, then evaluated on an RI-only test set to check sample quality. The generated continuations are attached to the real RI data points and added to the training set for the final classification module, another multivariate LSTM with a softmax output over RI versus non-RI. The design is what makes the oversampling spatiotemporal: unlike SMOTE, it does not interpolate pointwise, and unlike image-generation GANs, it produces continuous tracks with associated wind intensities.

What would settle it

Inspect the released code and data-preparation steps and compare the list of cyclones used to train the augmentation generator with the cyclones in the final held-out test set; if any RI cyclone appears in both, retrain DA-M-LSTM with the generator restricted to cyclones ending before the test period begins and remeasure the F1 scores.

Watch

Extended reading notes

Core claim

The central claim is that an LSTM can act as a spatiotemporal data augmentation engine for a class-imbalance problem: it learns the joint distribution of latitude, longitude, and wind intensity along RI cyclone tracks and generates synthetic four-point continuations that, when appended to real RI segments, substantially improve detection by the same classifier family. The paper reports that the augmented model DA-M-LSTM achieves an RI F1 of 0.651 ± 0.036 in the South Pacific and 0.657 ± 0.044 in the South Indian Ocean, compared with 0.339 ± 0.053 and 0.487 ± 0.029 for the multivariate LSTM baseline on the same test sets. The authors interpret this as evidence that synthetic oversampling, not just a larger training set, is what drives the gain, and that track coordinates carry signal that wind intensity alone misses.

Load-bearing premise

The load-bearing premise is that the synthetic RI samples used to augment the training set come from a generator whose training and test data are both strictly separated in time from the final classifier's held-out test period; the paper only says the split was 'based on period' and never states how the generator's RI-only test set relates to that held-out set, so leakage there would inflate the reported improvement.

Editorial extensions

If this is right

  • If the reported gains are reproducible, RI detection in data-sparse basins can improve without new observations, by generating synthetic RI tracks for training.
  • The framework makes spatial coordinates (latitude and longitude) a required input feature for RI detection models, not an optional extra.
  • The same LSTM-generator-plus-classifier recipe can be carried over to other rare spatiotemporal events, such as sudden flood onset or extreme heat waves, where the minority class is similarly underrepresented.
  • The reported precision, recall, and F1 benchmarks across U-LSTM, M-LSTM, E-LSTM, HE-LSTM, and DA-M-LSTM give future RI detection studies a direct comparison table.
  • Operational forecasting centers could use the augmented classifier as a screening tool to flag cyclones likely to undergo RI within the next 24 hours.

Reading between the lines

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

  • Not stated in the paper: the same generator could be run forward at forecast time to produce an ensemble of plausible RI track continuations, giving forecasters scenario diversity rather than only a classification label.
  • Not stated in the paper: the current experiments do not directly compare DA-M-LSTM against SMOTE or a GAN baseline on these basins, so a fair test would quantify how much of the gain comes specifically from preserving spatial continuity.
  • Not stated in the paper: a simple ablation varying how many synthetic four-point segments are added per real RI segment would reveal whether the improvement saturates or degrades, clarifying whether the method is regularizing the classifier or merely rebalancing the classes.
  • Not stated in the paper: if the temporal split is clean, the results suggest that track geometry itself is a discriminative cue for RI, which could be tested by feeding the same models only the displacement vectors between successive points rather than raw coordinates.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. This paper addresses class imbalance in cyclone rapid intensification (RI) detection by training LSTM models on wind intensity and spatial coordinates from JTWC best-track data for two Southern Hemisphere basins (1980-2020). The authors compare univariate, multivariate, ensemble, and hybrid LSTM classifiers and introduce an LSTM-based data augmentation module that generates four future time steps of latitude, longitude, and wind intensity for RI cyclone instances. Augmented synthetic samples are added to the training set, and the final DA-M-LSTM model is evaluated on the original real test set. The main reported results are large improvements in RI F1-score with data augmentation (e.g., South Pacific 0.339 to 0.651; South Indian 0.487 to 0.657 in Table 6). The abstract also claims spatial coordinates are critical input features.

Significance. If the reported improvements are not artifacts of test-set contamination or of simple class oversampling, the paper would provide a useful demonstration that sequence-generating LSTMs can mitigate class imbalance for spatiotemporal extreme-event detection. The study has methodological strengths: 30 independent runs with 95% confidence intervals, availability of code and data, and a clearly defined RI definition. The non-overlapping confidence intervals in Table 6 make the DA-M-LSTM gain statistically plausible. However, the paper's central attribution of the gain to the LSTM generator is not yet supported by the experiments as described, because the temporal separation in the augmentation pipeline is not documented and no simpler oversampling baseline is provided. The spatial-coordinates claim is also overstated relative to the benchmark tables.

major comments (3)
  1. [§3.3.2, §3.5, Table 2] The temporal separation between the augmentation generator's test set and the classifier's hold-out is not established. Table 2 reports 'Time-frame 1980-2020' for both the training and test partitions, while §3.3.2 states that the augmentation LSTM is evaluated on a test set containing only RI cyclones and that the generated continuations are added to the training set of the original dataset. If this RI-only test set is the same as, or overlaps with, the classifier's final evaluation period, then real prefixes from test RI events enter the augmented training data and the Table 6 F1 improvements (e.g., 0.339 to 0.651 for South Pacific) would be inflated by leakage. The authors must state the actual cutoff dates for the augmentation and classification splits and, if necessary, re-run the experiments with strict temporal isolation.
  2. [§4.3, Table 6] No baseline that oversamples the minority class by a simpler mechanism is provided. The DA-M-LSTM model differs from M-LSTM both in the presence of synthetic samples and in the total number of RI training instances; without a comparison to M-LSTM trained with duplicated RI samples (or a SMOTE-style time-series interpolation), the observed F1 gain cannot be attributed specifically to the LSTM generator's ability to produce realistic spatiotemporal continuations rather than to any oversampling. Please add such a baseline and state the final class ratio in the augmented training set.
  3. [Abstract, §4.2, Tables 3-5] The claim that spatial coordinates play a critical role is not supported by the reported results. For the South Indian basin at n=6 (Table 3), U-LSTM achieves RI F1=0.4798 versus M-LSTM's 0.4336; in Table 4 (South Pacific) the difference is 0.3368 versus 0.339; in Table 5 (South Indian) it is 0.4825 versus 0.4866. In every case the differences are small relative to the reported 95% confidence intervals. Either provide a statistical comparison (e.g., paired tests over the 30 runs) or temper the abstract's claim about the critical role of spatial coordinates.
minor comments (6)
  1. [§3.4, after Eq. (5)] The stray word 'begin' appears in the text, likely a LaTeX artifact; please remove it.
  2. [§3.3.1] The phrase 'data argumentation' should be 'data augmentation'.
  3. [§4.3] Please fix the typos 'generatd' (should be 'generated') and 'access the results' (should be 'assess the results').
  4. [§5] The phrase 'data from other courses' should likely be 'data from other sources'.
  5. [Table 2] The 'Time-frame' column is confusing; as printed it suggests both train and test span 1980-2020, which contradicts the description of a chronological 75/25 split in §3.5. Please clarify the actual date ranges for each partition.
  6. [References] Reference [77] lacks full bibliographic details; please update it to a complete citation.

Circularity Check

1 steps flagged · score 7.0 of 10

DA-M-LSTM gain is forced: generated continuations of test-set RI cyclones are added to the training set, so the reported improvement is by construction.

  1. fitted input called prediction [Section 3.3.2 (Data augmentation framework), Section 3.5 (Experiment setup), Table 2]
    "in the test set, we had only RI cyclones, so that later we can add the newly generated 4 points to the original 4 data points of RI instances, later added to train set of the original dataset, for training the classification model. ... We use 75 percent data in the training set and 25 percent data in the test set, where the data was separated based on period. ... Table 2: Train 1980-2020 / Test 1980-2020 for both oceans."

    The augmentation generator is applied to 'the test set' of RI cyclones, and its synthetic continuations, together with the original 4-point prefixes of those test RI instances, are inserted into the classification model's training set. The classifier is then evaluated on the same 25% test split: Section 3.5 defines one train/test split 'separated based on period', and Table 2 lists both Train and Test as 1980-2020, so no temporal separation of the augmentation test set from the classifier hold-out is shown. Consequently, DA-M-LSTM is trained on continuations of the very test RI tracks it is asked to predict, making the Table 6 improvement (e.g., South Pacific RI F1 0.339 -> 0.651) a forced in-sample result rather than an independent out-of-sample prediction.

full rationale

The central claim is that data augmentation improves rapid-intensification detection. The load-bearing step is the provenance of the synthetic augmentation samples. As written, Section 3.3.2 says the augmentation LSTM is tested on an RI-only 'test set' and that the generated 4 points plus the original 4 points of those RI instances are later added to the classifier's training set. Since Section 3.5 describes a single 75/25 train/test split 'separated based on period' and Table 2 gives the same 1980-2020 timeframe for both Train and Test, the augmentation samples appear to be derived from the classifier's own test set. The classifier is then evaluated on that same set, so the DA-M-LSTM result is contaminated by construction. This is the only significant circularity: the 'prediction' of test RI events is not independent of the training data. Self-citations (e.g., Chandra et al. on coevolutionary RNNs and cyclone modelling) are motivational and not load-bearing; no uniqueness theorem or ansatz is smuggled in via self-citation, and the core LSTM architecture is standard. If a future version explicitly shows that the augmentation test set is temporally disjoint from the classifier hold-out, the circularity would disappear and the score would drop to 0-2; but the current text does not establish that separation.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim depends on several hand-chosen hyperparameters and domain assumptions. No new physical entities are introduced. The most consequential choices are the post hoc selection of n=6 and the unverified realism of LSTM-generated synthetic tracks.

free parameters (5)
  • input time steps n = 6
    Selected as optimal based on F1 score in Table 3 across n = 5,6,7,8, a post hoc choice on test data.
  • hidden neurons = 50
    Chosen by trial experimental runs (Section 3.5, Table 1).
  • training epochs = 100
    Chosen by trial experimental runs (Table 1).
  • augmentation generator input/output length = 4 time steps
    The generator uses 4 past steps to predict 4 future steps; no justification given for this length.
  • learning rate = not reported
    Adam optimizer used but learning rate is not stated, an unlisted hand-choice that affects all results.
assumptions (3)
  • domain assumption JTWC best-track data for the Southern Hemisphere 1980-2020 are accurate and temporally consistent.
    All experiments use this data without independent verification (Section 3.1).
  • domain assumption Rapid intensification is adequately defined by a 30-knot increase in wind intensity within 24 hours.
    Standard NHC/JTWC definition adopted from literature (Section 1).
  • domain assumption An LSTM trained to forecast 4 future time steps of track and wind can generate synthetic sequences representative of real RI events.
    The augmentation framework in Section 3.3.2 relies on this to create useful training data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatiotemporal deep learning models for detection of rapid intensification in cyclones." pith.science (2026). https://pith.science/paper/VN3WKTPV

@misc{pith2026250608397,
  author       = {Pith},
  title        = {Pith review of: Spatiotemporal deep learning models for detection of rapid intensification in cyclones},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VN3WKTPV}},
  note         = {Machine review of arXiv:2506.08397}
}
read the original abstract

Cyclone rapid intensification is the rapid increase in cyclone wind intensity, exceeding a threshold of 30 knots, within 24 hours. Rapid intensification is considered an extreme event during a cyclone, and its occurrence is relatively rare, contributing to a class imbalance in the dataset. A diverse array of factors influences the likelihood of a cyclone undergoing rapid intensification, further complicating the task for conventional machine learning models. In this paper, we evaluate deep learning, ensemble learning and data augmentation frameworks to detect cyclone rapid intensification based on wind intensity and spatial coordinates. We note that conventional data augmentation methods cannot be utilised for generating spatiotemporal patterns replicating cyclones that undergo rapid intensification. Therefore, our framework employs deep learning models to generate spatial coordinates and wind intensity that replicate cyclones to address the class imbalance problem of rapid intensification. We also use a deep learning model for the classification module within the data augmentation framework to differentiate between rapid and non-rapid intensification events during a cyclone. Our results show that data augmentation improves the results for rapid intensification detection in cyclones, and spatial coordinates play a critical role as input features to the given models. This paves the way for research in synthetic data generation for spatiotemporal data with extreme events.

Figures

Figures reproduced from arXiv: 2506.08397 by the authors.

Figure 1
Figure 1. Spatio-temporal tracks of cyclones in the South Indian Ocean basin from 2009 to 2019, highlighting instances of Rapid Intensification (RI) and non-RI [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Spatio-temporal tracks of cyclones in the South Indian Ocean basin, categorized by wind speed-based classifications. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. LSTM model configurations for predicting rapid intensification of cyclones. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Proposed framework for cyclone RI prediction: The process involves data cleaning and RI case extraction, followed by LSTM-based data augmentation [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the data augmentation strategy using an LSTM model. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Life cycle of Winston cyclone in the South Pacific Ocean, highlighting a 24-hour window where RI event is detected (greater than or equal to 30 knots). [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Cyclone frequency over the years for the respective datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Frequency of RI events for the respective datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Number of RI events within each cyclone category for the two oceans. [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Histograms comparing the average wind intensity of original and augmented Rapid Intensification (RI) cyclone instances in the South Pacific and South [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Plots of tracks of synthetically generated cyclone instances [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: Cyclone tracks showing RI points detected and those missed by DA-M-LSTM [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

112 extracted references · 72 canonical work pages

  1. [1]

    Peduzzi, B

    P. Peduzzi, B. Chatenoux, H. Dao, A. De Bono, C. Herold, J. Kossin, F. Mouton, O. Nordbeck, Global trends in tropical cyclone risk, Nature climate change 2 (2012) 289–294

    2012

  2. [2]

    Miyamoto, T

    Y . Miyamoto, T. Takemi, A triggering mechanism for rapid intensifica- tion of tropical cyclones, Journal of the Atmospheric Sciences 72 (2015) 2666–2681

    2015

  3. [3]

    M. F. Wehner, C. Zarzycki, C. Patricola, Estimating the human influence on tropical cyclone intensity as the climate changes, in: Hurricane Risk, Springer, 2019, pp. 235–260

    2019

  4. [4]

    E. N. Rappaport, J.-G. Jiing, C. W. Landsea, S. T. Murillo, J. L. Franklin, The joint hurricane test bed: Its first decade of tropical cyclone research- to-operations activities reviewed, Bulletin of the American Meteorolog- ical Society 93 (2012) 371–380

    2012

  5. [5]

    U. S. N. Oceanic, A. A. N. W. Service, National hurricane center, glossary of nhc terms, 2013. URL: https://www.nhc.noaa.gov/ aboutgloss.shtml. 2https://github.com/sydney-machine-learning/cyclone_ deeplearning 14

    2013

  6. [6]

    K. T. Bhatia, G. A. Vecchi, T. R. Knutson, H. Murakami, J. Kossin, K. W. Dixon, C. E. Whitlock, Recent increases in tropical cyclone in- tensification rates, Nature communications 10 (2019) 1–9

    2019

  7. [7]

    Kaplan, C

    J. Kaplan, C. M. Rozo ff, M. DeMaria, C. R. Sampson, J. P. Kossin, C. S. Velden, J. J. Cione, J. P. Dunion, J. A. Kna ff, J. A. Zhang, et al., Eval- uating environmental impacts on tropical cyclone rapid intensification predictability utilizing statistical models, Weather and Forecasting 30 (2015) 1374–1396

    2015

  8. [8]

    Molinari, D

    J. Molinari, D. V ollaro, External influences on hurricane intensity. part ii: Vertical structure and response of the hurricane vortex, Journal of the Atmospheric sciences 47 (1990) 1902–1918

    1990

Show all 112 references
  1. [9]

    Kaplan, M

    J. Kaplan, M. DeMaria, J. A. Kna ff, A revised tropical cyclone rapid intensification index for the atlantic and eastern north pacific basins, Weather and forecasting 25 (2010) 220–241

    2010

  2. [10]

    J. J. Cione, E. W. Uhlhorn, Sea surface temperature variability in hurri- canes: Implications with respect to intensity change, Monthly Weather Review 131 (2003) 1783–1796

    2003

  3. [11]

    R. F. Rogers, P. D. Reasor, J. A. Zhang, Multiscale structure and evo- lution of hurricane earl (2010) during rapid intensification, Monthly Weather Review 143 (2015) 536–562

    2010

  4. [12]

    R. Chen, W. Zhang, X. Wang, Machine learning in tropical cyclone forecast modeling: A review, Atmosphere 11 (2020) 676

    2020

  5. [13]

    R. Chen, X. Wang, W. Zhang, X. Zhu, A. Li, C. Yang, A hybrid cnn-lstm model for typhoon formation forecasting, GeoInformatica 23 (2019) 375–396

    2019

  6. [14]

    J. Lian, P. Dong, Y . Zhang, J. Pan, A novel deep learning approach for tropical cyclone track prediction based on auto-encoder and gated recurrent unit networks, Applied Sciences 10 (2020) 3965

    2020

  7. [15]

    S. Yuan, C. Wang, B. Mu, F. Zhou, W. Duan, Typhoon intensity fore- casting based on lstm using the rolling forecast method, Algorithms 14 (2021) 83

    2021

  8. [16]

    Hochreiter, J

    S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural com- putation 9 (1997) 1735–1780

    1997

  9. [17]

    X. Wang, W. Wang, B. Yan, Tropical cyclone intensity change predic- tion based on surrounding environmental conditions with deep learning, Water 12 (2020) 2685

    2020

  10. [18]

    Varalakshmi, N

    P. Varalakshmi, N. Vasumathi, R. Venkatesan, Tropical cyclone predic- tion based on multi-model fusion across indian coastal region, Progress in Oceanography 193 (2021) 102557

    2021

  11. [19]

    Wimmers, C

    A. Wimmers, C. Velden, J. H. Cossuth, Using deep learning to es- timate tropical cyclone intensity from satellite passive microwave im- agery, Monthly Weather Review 147 (2019) 2261–2282

    2019

  12. [20]

    Maskey, R

    M. Maskey, R. Ramachandran, M. Ramasubramanian, I. Gurung, B. Fre- itag, A. Kaulfus, D. Bollinger, D. J. Cecil, J. Miller, Deepti: Deep- learning-based tropical cyclone intensity estimation system, IEEE Jour- nal of Selected Topics in Applied Earth Observations and Remote Sens...

    2020

  13. [21]

    Rajesh, V

    K. Rajesh, V . Ramaswamy, K. Kannan, Prediction of cyclone us- ing kalman spatio temporal and two dimensional deep learning model, Malaysian Journal of Computer Science (2020) 24–38

    2020

  14. [22]

    J. Lee, J. Im, D.-H. Cha, H. Park, S. Sim, Tropical cyclone intensity estimation using multi-dimensional convolutional neural networks from geostationary satellite data, Remote Sensing 12 (2020) 108

    2020

  15. [23]

    Wei, Study on wind simulations using deep learning techniques during typhoons: a case study of northern taiwan, Atmosphere 10 (2019) 684

    C.-C. Wei, Study on wind simulations using deep learning techniques during typhoons: a case study of northern taiwan, Atmosphere 10 (2019) 684

    2019

  16. [24]

    K. A. Cloud, B. J. Reich, C. M. Rozo ff, S. Alessandrini, W. E. Lewis, L. Delle Monache, A feed forward neural network based on model out- put statistics for short-term hurricane intensity prediction, Weather and Forecasting 34 (2019) 985–997

    2019

  17. [25]

    Chandra, K

    R. Chandra, K. Dayal, N. Rollings, Application of cooperative neuro- evolution of Elman recurrent networks for a two-dimensional cyclone track prediction for the South Pacific region, in: International Joint Conference on Neural Networks (IJCNN), 2015, pp. 1–8. doi:10.1109/ IJ...

    2015

  18. [26]

    Chandra, R

    R. Chandra, R. Deo, C. W. Omlin, An architecture for encoding two- dimensional cyclone track prediction problem in coevolutionary recur- rent neural networks, in: International Joint Conference on Neural Net- works (IJCNN), IEEE, 2016, pp. 4865–4872

    2016

  19. [27]

    Zhang, R

    Y . Zhang, R. Chandra, J. Gao, Cyclone track prediction with matrix neu- ral networks, in: International Joint Conference on Neural Networks, Rio de Janeiro, Brazil, IEEE, 2018, pp. 1–8. doi: 10.1109/IJCNN. 2018.8489077

    2018

  20. [28]

    Chandra, Y

    R. Chandra, Y . Ong, C. Goh, Co-evolutionary multi-task learning for dynamic time series prediction, Applied Soft Computing 70 (2018) 576– 589

    2018

  21. [29]

    Chandra, Dynamic cyclone wind-intensity prediction using co- evolutionary multi-task learning, in: D

    R. Chandra, Dynamic cyclone wind-intensity prediction using co- evolutionary multi-task learning, in: D. Liu, S. Xie, Y . Li, D. Zhao, E. M. El-Alfy (Eds.), International Conference on Neural Information Processing, Guangzhou, China, volume 10638 ofLecture Notes in Com- puter ...

    2017

  22. [30]

    Kapoor, A

    A. Kapoor, A. Negi, L. Marshall, R. Chandra, Cyclone trajectory and in- tensity prediction with uncertainty quantification using variational recur- rent neural networks, Environmental Modelling & Software 162 (2023) 105654

    2023

  23. [31]

    M. S. Fischer, B. H. Tang, K. L. Corbosiero, A climatological anal- ysis of tropical cyclone rapid intensification in environments of upper- tropospheric troughs, Monthly Weather Review 147 (2019) 3693–3719

    2019

  24. [32]

    R. Lu, X. Tang, Relationship between early-stage features and lifetime maximum intensity of tropical cyclones over the western north pacific, Atmosphere 12 (2021) 815

    2021

  25. [33]

    H. Su, L. Wu, J. H. Jiang, R. Pai, A. Liu, A. J. Zhai, P. Tavallali, M. De- Maria, Applying satellite observations of tropical cyclone internal struc- tures to rapid intensification forecast with machine learning, Geophysi- cal Research Letters 47 (2020) e2020GL089102

    2020

  26. [34]

    Yang, C.-Y

    Q. Yang, C.-Y . Lee, M. K. Tippett, A long short-term memory model for global rapid intensification prediction, Weather and Forecasting 35 (2020) 1203–1220

    2020

  27. [35]

    Chandra, K

    R. Chandra, K. S. Dayal, Coevolutionary recurrent neural networks for prediction of rapid intensification in wind intensity of tropical cyclones in the south pacific region, 2015

    2015

  28. [36]

    Chandra, Towards prediction of rapid intensification in tropical cy- clones with recurrent neural networks, arXiv preprint arXiv:1701.04518 (2017)

    R. Chandra, Towards prediction of rapid intensification in tropical cy- clones with recurrent neural networks, arXiv preprint arXiv:1701.04518 (2017)

    2017 arXiv

  29. [37]

    N. V . Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer, Smote: synthetic minority over-sampling technique, Journal of artificial intelli- gence research 16 (2002) 321–357

    2002

  30. [38]

    Torgo, R

    L. Torgo, R. P. Ribeiro, B. Pfahringer, P. Branco, SMOTE for regression, in: Portuguese conference on artificial intelligence, Springer, 2013, pp. 378–389

    2013

  31. [39]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde- Farley, S. Ozair, A. Courville, Y . Bengio, Generative ad- versarial nets, in: Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, K. Q. Weinberger (Eds.), Advances in Neural In- formation Processing Systems, volume 27...

    2014

  32. [40]

    M. Buda, A. Maki, M. A. Mazurowski, A systematic study of the class imbalance problem in convolutional neural networks, Neural Networks 106 (2018) 249–259

    2018

  33. [41]

    Galar, A

    M. Galar, A. Fernandez, E. Barrenechea, H. Bustince, F. Herrera, A re- view on ensembles for the class imbalance problem: bagging-, boosting- , and hybrid-based approaches, IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 42 (2011) 463–484

    2011

  34. [42]

    S. M. Abd Elrahman, A. Abraham, A review of class imbalance prob- lem, Journal of Network and Innovative Computing 1 (2013) 332–340

    2013

  35. [43]

    Gosain, S

    A. Gosain, S. Sardana, Handling class imbalance problem using over- sampling techniques: A review, in: 2017 international conference on ad- vances in computing, communications and informatics (ICACCI), IEEE, 2017, pp. 79–85

    2017

  36. [44]

    Y . Wei, R. Yang, An advanced artificial intelligence system for inves- tigating tropical cyclone rapid intensification with the ships database, Atmosphere 12 (2021) 484

    2021

  37. [45]

    S. S. Chand, K. J. Tory, H. Ye, K. J. Walsh, Projected increase in el ni˜no-driven tropical cyclone frequency in the pacific, Nature Climate Change 7 (2017) 123–127

    2017

  38. [46]

    B. S. Raghuwanshi, S. Shukla, Smote based class-specific extreme learning machine for imbalanced learning, Knowledge-Based Systems 187 (2020) 104814

    2020

  39. [47]

    J. B. Elsner, J. P. Kossin, T. H. Jagger, The increasing intensity of the strongest tropical cyclones, Nature 455 (2008) 92–95

    2008

  40. [48]

    J. M. Johnson, T. M. Khoshgoftaar, Survey on deep learning with class 15 imbalance, Journal of Big Data 6 (2019) 1–54

    2019

  41. [49]

    O. S. Sitompul, E. B. Nababan, et al., Biased support vector machine and weighted-smote in handling class imbalance problem, International Journal of Advances in Intelligent Informatics 4 (2018) 21–27

    2018

  42. [50]

    Y . Xu, H. Yang, M. Cheng, S. Li, Cyclone intensity estimate with context-aware cyclegan, in: 2019 IEEE International Conference on Image Processing (ICIP), IEEE, 2019, pp. 3417–3421. doi: 10.1109/ ICIP.2019.8803598

    2019

  43. [51]

    J. Yan, S. Han, Classifying imbalanced data sets by a novel re-sample and cost-sensitive stacked generalization method, Mathematical Prob- lems in Engineering 2018 (2018)

    2018

  44. [52]

    S. Wang, Z. Li, W. Chao, Q. Cao, Applying adaptive over-sampling technique based on data density and cost-sensitive svm to imbalanced learning, in: The 2012 International Joint Conference on Neural Net- works (IJCNN), IEEE, 2012, pp. 1–8

    2012

  45. [53]

    A. A. Khan, O. Chaudhari, R. Chandra, A review of ensemble learning and data augmentation models for class imbalanced problems: Combi- nation, implementation and evaluation, Expert Systems with Applica- tions 244 (2024) 122778

    2024

  46. [54]

    Fern ´andez, S

    A. Fern ´andez, S. Garcia, F. Herrera, N. V . Chawla, Smote for learning from imbalanced data: progress and challenges, marking the 15-year anniversary, Journal of artificial intelligence research 61 (2018) 863– 905

    2018

  47. [55]

    W. Wang, S. Wang, W. Fan, Z. Liu, J. Tang, Global-and-local aware data generation for the class imbalance problem, in: Proceedings of the 2020 SIAM International Conference on Data Mining, SIAM, 2020, pp. 307–315

    2020

  48. [56]

    Ali-Gombe, E

    A. Ali-Gombe, E. Elyan, MFC-GAN: class-imbalanced dataset classifi- cation using multiple fake class generative adversarial network, Neuro- computing 361 (2019) 212–221

    2019

  49. [57]

    Mariani, F

    G. Mariani, F. Scheidegger, R. Istrate, C. Bekas, C. Malossi, BAGAN: Data augmentation with balancing GAN, arXiv preprint arXiv:1803.09655 (2018)

    2018 arXiv

  50. [58]

    Sharma, P

    A. Sharma, P. K. Singh, R. Chandra, Smotified-gan for class imbal- anced pattern classification problems, arXiv preprint arXiv:2108.03235 (2021)

    2021 arXiv

  51. [59]

    Creswell, T

    A. Creswell, T. White, V . Dumoulin, K. Arulkumaran, B. Sengupta, A. A. Bharath, Generative adversarial networks: An overview, IEEE signal processing magazine 35 (2018) 53–65

    2018

  52. [60]

    J. Gui, Z. Sun, Y . Wen, D. Tao, J. Ye, A review on generative adversarial networks: Algorithms, theory, and applications, IEEE Transactions on Knowledge and Data Engineering (2021)

    2021

  53. [61]

    X. Yi, E. Walia, P. Babyn, Generative adversarial network in medical imaging: A review, Medical image analysis 58 (2019) 101552

    2019

  54. [62]

    Scott, J

    M. Scott, J. Plested, Gan-smote: A generative adversarial network ap- proach to synthetic minority oversampling., Aust. J. Intell. Inf. Process. Syst. 15 (2019) 29–35

    2019

  55. [63]

    N. M. Nafi, W. H. Hsu, Addressing class imbalance in image-based plant disease detection: Deep generative vs. sampling-based approaches, in: 2020 International Conference on Systems, Signals and Image Process- ing (IWSSIP), IEEE, 2020, pp. 243–248

    2020

  56. [64]

    T. R. Knutson, J. L. McBride, J. Chan, K. Emanuel, G. Holland, C. Land- sea, I. Held, J. P. Kossin, A. Srivastava, M. Sugi, Tropical cyclones and climate change, Nature geoscience 3 (2010) 157–163

    2010

  57. [65]

    Murakami, T

    H. Murakami, T. L. Delworth, W. F. Cooke, M. Zhao, B. Xiang, P.- C. Hsu, Detected climatic change in global distribution of tropical cy- clones, Proceedings of the National Academy of Sciences 117 (2020) 10706–10714

    2020

  58. [66]

    Tauvale, K

    L. Tauvale, K. Tsuboki, Characteristics of tropical cyclones in the south- west pacific, Journal of the Meteorological Society of Japan. Ser. II (2019)

    2019

  59. [67]

    Magee, D

    A. Magee, D. Verdon-Kidd, Historical variability of southwest pacific tropical cyclone counts since 1855, Geophysical Research Letters 46 (2019) 6936–6945

    2019

  60. [68]

    A. D. Magee, D. C. Verdon-Kidd, H. J. Diamond, A. S. Kiem, Influence of enso, enso modoki, and the ipo on tropical cyclogenesis: a spatial analysis of the southwest pacific region, International Journal of Clima- tology 37 (2017) 1118–1137

    2017

  61. [69]

    McInnes, J

    K. McInnes, J. O’Grady, K. Walsh, F. Colberg, Progress towards quanti- fying storm surge risk in fiji due to climate variability and change, Jour- nal of Coastal Research (2011) 1121–1124

    2011

  62. [70]

    KOSTASCHUK, J

    R. KOSTASCHUK, J. TERRY , R. RAJ, Tropical cyclones and floods in fiji, Hydrological Sciences Journal 46 (2001) 435–450

    2001

  63. [71]

    E. K. Pradhan, K. P. West Jr, J. Katz, S. C. LeClerq, S. K. Khatry, S. R. Shrestha, Risk of flood-related mortality in nepal, Disasters 31 (2007) 57–70

    2007

  64. [72]

    Madsen, F

    H. Madsen, F. Jakobsen, Cyclone induced storm surge and flood fore- casting in the northern bay of bengal, Coastal Engineering 51 (2004) 277–296

    2004

  65. [73]

    J. P. Terry, Emerging concepts in hydrology for tropical pacific regimes, in: Advances in Geosciences: V olume 11: Hydrological Science (HS), World Scientific, 2009, pp. 45–55

    2009

  66. [74]

    Alexander, A magnitude scale for cascading disasters, International journal of disaster risk reduction 30 (2018) 180–185

    D. Alexander, A magnitude scale for cascading disasters, International journal of disaster risk reduction 30 (2018) 180–185

    2018

  67. [75]

    Munshi, A general linear model for trends in tropical cyclone activity, Available at SSRN 2630932 (2015)

    J. Munshi, A general linear model for trends in tropical cyclone activity, Available at SSRN 2630932 (2015)

    2015

  68. [76]

    S. C. Steenkamp, G. Kilroy, R. K. Smith, Tropical cyclogenesis at and near the equator, Quarterly Journal of the Royal Meteorological Society 145 (2019) 1846–1864

    2019

  69. [77]

    J. T. W. C. (JTWC), The tropical cyclones best track data in the southern hemisphere (1980-2020)

    1980

  70. [78]

    LeCun, Y

    Y . LeCun, Y . Bengio, G. Hinton, Deep learning, nature 521 (2015) 436–444

    2015

  71. [79]

    Y . Guo, Y . Liu, A. Oerlemans, S. Lao, S. Wu, M. S. Lew, Deep learning for visual understanding: A review, Neurocomputing 187 (2016) 27–48

    2016

  72. [80]

    Scher, Toward data-driven weather and climate forecasting: Approxi- mating a simple general circulation model with deep learning, Geophys- ical Research Letters 45 (2018) 12–616

    S. Scher, Toward data-driven weather and climate forecasting: Approxi- mating a simple general circulation model with deep learning, Geophys- ical Research Letters 45 (2018) 12–616

    2018

  73. [81]

    Ardabili, A

    S. Ardabili, A. Mosavi, M. Dehghani, A. R. V ´arkonyi-K´oczy, Deep learning and machine learning in hydrological processes climate change and earth systems a systematic review, in: International conference on global research and education, Springer, 2019, pp. 52–62

    2019

  74. [82]

    P. Wang, E. Fan, P. Wang, Comparative analysis of image classification algorithms based on traditional machine learning and deep learning, Pat- tern Recognition Letters 141 (2021) 61–67

    2021

  75. [83]

    Roscher, B

    R. Roscher, B. Bohn, M. F. Duarte, J. Garcke, Explainable machine learning for scientific insights and discoveries, IEEE Access 8 (2020) 42200–42216

    2020

  76. [84]

    J. L. Elman, Finding structure in time, Cognitive science 14 (1990) 179–211

    1990

  77. [85]

    C. L. Giles, C. W. Omlin, K. K. Thornber, Equivalence in knowl- edge representation: automata, recurrent neural networks, and dynami- cal fuzzy systems, Proceedings of the IEEE 87 (1999) 1623–1640

    1999

  78. [86]

    P. J. Werbos, Backpropagation through time: what it does and how to do it, Proceedings of the IEEE 78 (1990) 1550–1560

    1990

  79. [87]

    Bengio, P

    Y . Bengio, P. Simard, P. Frasconi, Learning long-term dependencies with gradient descent is difficult, IEEE transactions on neural networks 5 (1994) 157–166

    1994

  80. [88]

    Y . Bai, B. Zeng, C. Li, J. Zhang, An ensemble long short-term memory neural network for hourly pm2. 5 concentration forecasting, Chemo- sphere 222 (2019) 286–294

    2019

  81. [89]

    Mikolov, A

    T. Mikolov, A. Joulin, S. Chopra, M. Mathieu, M. Ranzato, Learn- ing longer memory in recurrent neural networks, arXiv preprint arXiv:1412.7753 (2014)

    2014 arXiv

  82. [90]

    Van Houdt, C

    G. Van Houdt, C. Mosquera, G. N ´apoles, A review on the long short- term memory model, Artificial Intelligence Review 53 (2020) 5929– 5955

    2020

  83. [91]

    Shi, D.-Y

    X. Shi, D.-Y . Yeung, Machine learning for spatiotemporal sequence forecasting: A survey, arXiv preprint arXiv:1808.06865 (2018)

    2018 arXiv

  84. [92]

    Y . Ding, Y . Zhu, J. Feng, P. Zhang, Z. Cheng, Interpretable spatio- temporal attention lstm model for flood forecasting, Neurocomputing 403 (2020) 348–359

    2020

  85. [93]

    L. Xu, K. C. Mo, A preliminary study of deep learning based drought forecast. climate prediction s&t digest, in: Science and Technology In- fusion Climate Bulletin NOAA’s National Weather Service, 44th NOAA Annual Climate Diagnostics and Prediction Workshop, 2019, pp. 22–24

    2019

  86. [94]

    Z. Ming, M. Visani, M. M. Luqman, J.-C. Burie, A survey on anti- spoofing methods for facial recognition with rgb cameras of generic con- sumer devices, Journal of Imaging 6 (2020) 139

    2020

  87. [95]

    Agrawal, D

    S. Agrawal, D. S. Sisodia, N. K. Nagwani, Long short term memory based functional characterization model for unknown protein sequences using ensemble of shallow and deep features, Neural Computing and 16 Applications (2021) 1–11

    2021

  88. [96]

    Chandra, S

    R. Chandra, S. Goyal, R. Gupta, Evaluation of deep learning models for multi-step ahead time series prediction, IEEE Access 9 (2021) 83105– 83123

    2021

  89. [97]

    Shorten, T

    C. Shorten, T. M. Khoshgoftaar, B. Furht, Text data augmentation for deep learning, Journal of big Data 8 (2021) 101

    2021

  90. [98]

    Shorten, T

    C. Shorten, T. M. Khoshgoftaar, A survey on image data augmentation for deep learning, Journal of big data 6 (2019) 1–48

    2019

  91. [99]

    Q. Wen, L. Sun, F. Yang, X. Song, J. Gao, X. Wang, H. Xu, Time series data augmentation for deep learning: A survey, arXiv preprint arXiv:2002.12478 (2020)

    2020 arXiv

  92. [100]

    D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  93. [101]

    Meyers, P

    G. Meyers, P. McIntosh, L. Pigot, M. Pook, The years of el ni˜no, la ni˜na, and interactions with the tropical indian ocean, Journal of Climate 20 (2007) 2872–2880

    2007

  94. [102]

    G. A. Vecchi, A. T. Wittenberg, El ni˜no and our future climate: where do we stand?, Wiley Interdisciplinary Reviews: Climate Change 1 (2010) 260–270

    2010

  95. [103]

    C. Mora, B. Dousset, I. R. Caldwell, F. E. Powell, R. C. Geronimo, C. R. Bielecki, C. W. W. Counsell, B. S. Dietrich, E. T. Johnston, L. V . Louis, et al., Extreme weather events exacerbated by anthropogenic climate change, Nature Climate Change 8 (2018) 106–113. doi: 10. 1038...

    2018

  96. [104]

    J. H. Faghmous, V . Kumar, A big data guide to understanding climate change: The case for theory-guided data science, Big data 2 (2014) 155–163

    2014

  97. [105]

    X. Chen, J. A. Zhang, F. D. Marks, A thermodynamic pathway leading to rapid intensification of tropical cyclones in shear, Geophysical Research Letters 46 (2019) 7717–7725. doi:10.1029/2019GL083667

    2019 doi

  98. [106]

    Sharma, P

    A. Sharma, P. K. Singh, R. Chandra, Smotified-gan for class imbalanced pattern classification problems, IEEE Access 10 (2022) 30655–30665

    2022

  99. [107]

    Goodfellow, Nips 2016 tutorial: generative adversarial networks

    I. Goodfellow, Nips 2016 tutorial: generative adversarial networks. corr abs/1701.00160, arXiv preprint arXiv:1701.00160 (2017)

    2017 arXiv

  100. [108]

    Arjovsky, S

    M. Arjovsky, S. Chintala, L. Bottou, Wasserstein generative adversarial networks, in: International conference on machine learning, PMLR, 2017, pp. 214–223

    2017

  101. [109]

    J. Li, A. Madry, J. Peebles, L. Schmidt, On the limitations of first- order approximation in gan dynamics, in: International Conference on Machine Learning, PMLR, 2018, pp. 3005–3013

    2018

  102. [110]

    G. Xu, Y . Li, C. Ma, X. Li, Y . Ye, Q. Lin, Z. Huang, S. Chen, TFG-Net: Tropical cyclone intensity estimation from a fine-grained perspective with the graph convolution neural network, Engineering Applications of Artificial Intelligence 118 (2023) 105673

    2023

  103. [111]

    Chandra, A

    R. Chandra, A. Bhagat, M. Maharana, P. N. Krivitsky, Bayesian graph convolutional neural networks via tempered MCMC, arXiv preprint arXiv:2104.08438 (2021)

    2021 arXiv

  104. [112]

    Chandra, R

    R. Chandra, R. Chen, J. Simmons, Bayesian neural networks via MCMC: a Python-based tutorial, IEEE Access (2024). 17

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.