Pith. sign in

REVIEW 4 major objections 5 minor 57 references

OASIS: Harnessing Diffusion Adversarial Network for Ocean Salinity Imputation using Sparse Drifter Trajectories

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A diffusion-adversarial network imputes coastal salinity from sparse drifter trajectories, using tidal height as a covariate and beating statistical and neural baselines on real and simulated data.

desk verdict Useful packaging of existing modules with a real deployment artifact, but the abstract overclaims consistency and the single-day real-world evaluation supports interpolation, not generalization. read the letter →

arxiv 2508.21570 v1 pith:CYZYCPNI submitted 2025-08-29 cs.LG

Bo Li , Yingqi Feng , Ming Jin , Xin Zheng , Yufei Tang , Laurent Cherubin , Alan Wee-Chung Liew , Can Wang
show 5 more authors
Qinghua Lu Jingwei Yao Shirui Pan Hong Zhang Xingquan Zhu
This is my paper · ORCID
classification cs.LG
keywords salinityimputationdriftertrajectoriesdiffusionadversarialnetworktransformerattentiontidalcovariatessparsedatacoastaloceangenerative
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

OASIS is a generative imputation system that reconstructs missing ocean salinity values from sparse, irregular drifter trajectories. The paper argues that a combination of per-trajectory normalization, a transformer module for long-range spatiotemporal dependencies, and a cosine-scheduled diffusion adversarial network conditioned on tidal height can outperform classical interpolation and neural baselines. On the real-world Fort Pierce Inlet dataset, OASIS reports a 21.3% lower RMSE and 18.5% lower MAPE than the best baseline (MLP), and on simulated Gulf of Mexico data it reports up to 52.5% lower MAE than Kriging. The authors also show through ablations that each architectural component contributes, with the global dependency capturing module giving the largest gain. If correct, the system offers a low-cost way to fill salinity gaps using only tide data that is already routinely observed.

What carries the argument

The central object is the Scheduler Diffusion Adversarial Network, a GAN in which a cosine noise schedule injects multi-scale noise into real and fake samples before the discriminator judges them, forcing the generator to refine imputed salinity fields progressively. The generator's loss combines mean squared error against observed salinity with a feature-matching term that aligns discriminator hidden activations for real and generated samples. This adversarial core is fed by a transformer-based global dependency capturing module with positional encoding and multi-head self-attention, so that each imputed cell can attend to every other cell in the spatiotemporal grid despite sparse coverage. Normalization and the tidal-height covariate condition the whole pipeline; the tide acts as a periodic proxy for physical drivers that would otherwise require specialized sensors.

What would settle it

Train OASIS on the Fort Pierce data from 8 December 2015 and 15 December 2015, test on the 16 June 2016 trajectories, and compare RMSE and MAPE with MLP and LSTM under the same settings; if OASIS no longer beats the baselines, the reported real-world gains are within-day interpolation rather than spatiotemporal generalization.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that ocean salinity imputation under severe drifter sparsity is best treated as a generative problem rather than a spatial interpolation or a forecasting problem. The proposed OASIS pipeline first normalizes each drifter trajectory with reversible instance normalization, then uses a transformer-based global dependency capturing module to encode long-range spatiotemporal correlations, and finally feeds the representation into a generator trained adversarially against a cosine-scheduled diffusion discriminator, all conditioned on tidal height as an easily observed proxy for unmeasured physical drivers. The authors report that OASIS consistently beats Kriging, geographically weighted regression, MLP, LSTM, and a vanilla GAN across one real dataset and three simulated Gulf of Mexico datasets, with the strongest margins on the real data. Ablations removing normalization, the global dependency module, or the diffusion scheduler all degrade performance, supporting the claim that each component is load-bearing.

Load-bearing premise

The real-world evaluation assumes that randomly splitting the four Fort Pierce trajectories recorded on a single day into train and test sets measures generalization to unseen conditions, even though all four trajectories sample the same small water body on the same day.

Editorial extensions

If this is right

  • Salinity maps can be produced at arbitrary times and locations from sparse drifter tracks, using only tidal height retrieved from routine tide records as an external input.
  • On the real-world data, the reported error reductions (21.3% RMSE, 18.5% MAPE over the best baseline) imply a practically meaningful accuracy gain for nearshore monitoring.
  • The ablation results imply that no single trick is responsible: normalization, global attention, and diffusion scheduling each contribute, and omitting the attention module costs the most.
  • Because the model trains on sea surface salinity only, the same deployment pipeline can be refreshed with new serialized model files without changing the interface.

Reading between the lines

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

  • A natural, stronger evaluation would split the real data by deployment day rather than by random trajectory, training on the 2015 days and testing on the 2016 day; if the advantage shrinks, part of the reported gain is interpolation inside one water mass.
  • The same conditioning trick could be tested with other periodic, easily observed covariates such as river stage or diurnal temperature, which would extend OASIS to estuaries where freshwater input, not tide, dominates salinity.
  • Because the method is formulated on a generic 4D spatiotemporal tensor, it could be applied to other sparse Lagrangian observations such as surface temperature, chlorophyll, or dissolved oxygen collected by drifters or floats.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces OASIS, a diffusion adversarial imputation model for sea surface salinity from sparse drifter trajectories. The architecture combines reversible instance normalization, a transformer-based global dependency module, and a GAN refined with cosine-scheduled diffusion; tidal height from a NOAA station is used as an auxiliary covariate. Evaluation is performed on one real-world day (Fort Pierce Inlet, four trajectories) and three simulated monthly Gulf of Mexico subsets, against Kriging, GWR, MLP, LSTM, and vanilla GAN. The authors report gains over Kriging and over MLP on the real data, and present ablations and a web deployment tool.

Significance. The application is timely and the system-building contribution is real: OASIS is one of the first end-to-end frameworks that combines drifters, tidal proxies, and generative modeling for salinity imputation, and the authors ship code and a lightweight web interface. If the empirical claims were robust, the work would be a useful addition to coastal ocean monitoring. However, the headline claims are not supported by the reported numbers: the 'consistent improvement' assertion is contradicted by Table 2, the abstract's 52.5% Kriging reduction cannot be traced, and the real-world evaluation does not demonstrate generalization beyond a single water mass. The significance of the result currently falls short of what the paper claims.

major comments (4)
  1. [Section 4.4, Table 2] The claim that 'OASIS consistently achieves superior performance' is contradicted by Table 2. On GoM-10, OASIS has MAPE 1.29% versus MLP's 1.15%; on GoM-11, MAPE 1.21% versus MLP's 1.03%; and on GoM-12, OASIS has MAE 0.4761 versus MLP's 0.3696, LSTM's 0.4679, and GAN's 0.4707, with MAPE 1.37% versus MLP's 1.07%. These are not isolated edge cases; they are four of the twelve dataset-metric cells. The paper's abstract, introduction, and conclusion should be revised to state the actual pattern, e.g., that OASIS wins on RMSE across the synthetic Gulf of Mexico sets and on the real-world Fort Pierce set, but not consistently on MAE or MAPE.
  2. [Section 4.3] The real-world evaluation uses a random 70/15/15 split with seed 42 applied to four trajectories from a single day (16 June 2016) within a 0.06-degree longitude by 0.01-degree latitude region (Table 1). Because test observations are spatiotemporally interleaved with training observations from the same water mass, the reported reductions (e.g., 21.3% RMSE over MLP) may reflect the model's ability to interpolate locally rather than to generalize to other times or locations. The paper needs a spatial or temporal holdout, multiple random seeds with variance estimates, or at least a discussion of this limitation. Without this, the 'robustness' conclusion in Section 4.4 is not supported.
  3. [Abstract] The abstract claims 'achieving up to 52.5% reduction in MAE compared to Kriging,' but this number does not appear anywhere in Table 2. Computing MAE reductions against Kriging from Table 2 gives 85.7% for FP Observed, 60.1% for GoM-10, 70.3% for GoM-11, and 68.7% for GoM-12. The 52.5% figure should either be reproduced from a specific comparison with a clear definition (e.g., RMSE on a particular split) or removed, since it is the headline quantitative claim of the paper.
  4. [Section 4.1] The Gulf of Mexico datasets are simulated drifter trajectories generated from a numerical ocean current model, not observational salinity fields. The paper does not describe how missingness is simulated for these sets or whether the evaluation on these sets corresponds to held-out trajectories, held-out timesteps, or artificially masked gauges. Without this information, the GoM results are difficult to interpret, especially because the baselines' relative performance varies by metric on these sets.
minor comments (5)
  1. [Table 3 caption] The caption says 'ablation on FB Observed dataset'; this should be 'FP Observed' to match the dataset name used throughout the paper.
  2. [Equation (7)] The attention output should be X_i = sum_j A_ij V_j; the current text writes V_i, which is inconsistent with the definition of attention as a convex combination of value vectors.
  3. [Equation (13)] Equation (13) uses q_t(G(X)) without defining q_t; please state how q_t relates to the noisy sample defined in Equation (8).
  4. [Section 4.5, Figure 3] The with-tide and without-tide comparison in Figure 3 is reported for FP Observed only, and no error bars or multiple-seed results are shown; please clarify whether the same train/validation/test split is used and whether the sinusoidal tide fit is evaluated on held-out tide data.
  5. [Section 4.2] The baseline set does not include a recent diffusion-based imputation method, even though OASIS is a diffusion-based model; adding a standard baseline such as CSDI would make the comparison more informative.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; OASIS is an empirical imputation system whose predictions are not equivalent to its inputs by construction.

full rationale

The paper's derivation chain is a standard supervised imputation pipeline: sparse salinity tensor X and mask M are the inputs; each drifter trajectory is normalized (Eqs. 2-3), spatiotemporal attention is applied (Eqs. 4-7), and a generator/discriminator pair is trained with reconstruction and feature-matching losses (Eqs. 9-13). The only externally fitted input is tidal height, which is constructed by fitting a sinusoid to NOAA tide observations in Section 4.5; tide is then used as a covariate to predict salinity, a different target, so no prediction reduces to a fitted parameter by construction. The same fixed seed-42 random split (Section 4.3) is applied to OASIS and all baselines, so the reported gains over MLP, LSTM, and GAN are not obtained by using test salinity during training. Author self-citations (e.g., refs [22, 50-53]) concern graph neural network and time-series methodology and are not load-bearing for the salinity imputation claim. The real-world evaluation limitation noted by the reader - a random split of four same-day, same-inlet trajectories - is a generalization-validity concern about the benchmark, not circularity: the model's definition and loss functions do not presuppose the result they are used to demonstrate. No equation in the paper is equivalent to its input by construction, and no fitted parameter is renamed as a prediction. Therefore, the appropriate circularity score is 0.

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

The central claim relies on data and evaluation assumptions: tide is a valid proxy, synthetic drifter data represent real fields, and a same-day random split is independent. No new physical entities are introduced; the model is a software system.

free parameters (3)
  • Diffusion scheduler hyperparameters (beta_0, beta_T, total steps T) = not reported
    Eq. 8 defines the cosine noise schedule using beta_0 and beta_T, but the paper does not give their values or the number of diffusion steps. These hand-chosen constants affect training and the final model.
  • OASIS architecture hyperparameters (layers, attention heads, hidden dimensions, learning rate, batch size) = not reported
    Section 3 specifies the modules but omits the concrete sizes and optimization settings needed to reproduce the reported numbers.
  • Tide sinusoid parameters (amplitude, phase, period) = fitted per station and date
    Section 4.5 fits a sinusoidal tide model from 2-4 NOAA tide events per day. The fitted tide level is used as a conditioning covariate in the generator.
assumptions (3)
  • domain assumption Tidal height is an easily observed proxy for the unmeasured physical drivers of nearshore salinity.
    The method conditions on tide as a proxy (abstract and Section 4.5), but the paper provides no physical validation that tidal height correlates with salinity in the Fort Pierce Inlet beyond the reported metric gains.
  • domain assumption Randomly splitting the four same-day FP Observed trajectories into train/validation/test sets yields independent test samples.
    Section 4.3 states trajectories are split 70/15/15 with seed 42. Because all trajectories are from 16 June 2016 in a small inlet, the split likely violates independence and inflates test performance.
  • domain assumption Simulated Gulf of Mexico drifter trajectories from the numerical ocean current model are representative of real nearshore salinity variability.
    Section 4.1 uses GoM simulated data as benchmarks. Performance on these data depends on the fidelity of the simulator, which is not assessed in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OASIS: Harnessing Diffusion Adversarial Network for Ocean Salinity Imputation using Sparse Drifter Trajectories." pith.science (2026). https://pith.science/paper/CYZYCPNI

@misc{pith2026250821570,
  author       = {Pith},
  title        = {Pith review of: OASIS: Harnessing Diffusion Adversarial Network for Ocean Salinity Imputation using Sparse Drifter Trajectories},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CYZYCPNI}},
  note         = {Machine review of arXiv:2508.21570}
}
read the original abstract

Ocean salinity plays a vital role in circulation, climate, and marine ecosystems, yet its measurement is often sparse, irregular, and noisy, especially in drifter-based datasets. Traditional approaches, such as remote sensing and optimal interpolation, rely on linearity and stationarity, and are limited by cloud cover, sensor drift, and low satellite revisit rates. While machine learning models offer flexibility, they often fail under severe sparsity and lack principled ways to incorporate physical covariates without specialized sensors. In this paper, we introduce the OceAn Salinity Imputation System (OASIS), a novel diffusion adversarial framework designed to address these challenges.

Figures

Figures reproduced from arXiv: 2508.21570 by the authors.

Figure 1
Figure 1. An overview of the two-stage pipeline for [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Comparison of sea surface salinity (psu) along [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison of different models on [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Overview of our OASIS: the web-based deployment interface (left) and the underlying system architecture (right). correlations is critical for accurate salinity imputation. (3) Without SD. Excluding the scheduler diffusion component also worsens re￾sults (MAE = 2.5280, …

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 48 canonical work pages

  1. [1]

    Laith Alzubaidi, Jinshuai Bai, Aiman Al-Sabaawi, Jose Santamaría, Ahmed Shihab Albahri, Bashar Sami Nayyef Al-Dabbagh, Mohammed A Fadhel, Mohamed Manoufali, Jinglan Zhang, Ali H Al-Timemy, et al . 2023. A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications. Journal of Big Data 10, 1 (2023), 46

    2023

  2. [2]

    Saqib Ejaz Awan, Mohammed Bennamoun, Ferdous Sohel, Frank Sanfilippo, and Girish Dwivedi. 2021. Imputation of missing data with class imbalance using conditional generative adversarial networks. Neurocomputing 453 (2021), 164–171

    work page 2021

  3. [3]

    Lorenzo Beretta and Alessandro Santaniello. 2016. Nearest neighbor imputation algorithms: a critical evaluation. BMC medical informatics and decision making 16 (2016), 197–208

    work page 2016

  4. [4]

    NA Bray. 1988. Water mass formation in the Gulf of California. Journal of Geophysical Research: Oceans 93, C8 (1988), 9223–9240

    work page 1988

  5. [5]

    Chris Brunsdon, A Stewart Fotheringham, and Martin E Charlton. 1996. Geo- graphically weighted regression: a method for exploring spatial nonstationarity. Geographical analysis 28, 4 (1996), 281–298

    work page 1996

  6. [6]

    Luca R Centurioni, Verena Hormann, Yi Chao, Gilles Reverdin, Jordi Font, and Dong-Kyu Lee. 2015. Sea surface salinity observations with Lagrangian drifters in the tropical North Atlantic during SPURS: Circulation, fluxes, and comparisons with remotely sensed salinity from Aquarius. Oceanography 28, 1 (2015), 96–105

    work page 2015

  7. [7]

    Astha Chawla, Prakhar Agrawal, Bijaya Ketan Panigrahi, and Kolin Paul. 2023. Deep-learning-based data-manipulation attack resilient supervisory backup pro- tection of transmission lines. Neural Computing and Applications 35, 7 (2023), 4835–4854

    work page 2023

  8. [8]

    Kai Chen, Minjie Ni, Minggang Cai, Jun Wang, Dongren Huang, Huorong Chen, Xiao Wang, and Mengyang Liu. 2016. Optimization of a coastal environmental monitoring network based on the Kriging method: A case study of Quanzhou Bay, China. BioMed research international 2016, 1 (2016), 7137310

    work page 2016

Show all 57 references
  1. [9]

    Eugenio Cutolo, Ananda Pascual, Simon Ruiz, Nikolaos D Zarokanellos, and Ronan Fablet. 2024. CLOINet: ocean state reconstructions through remote- sensing, in-situ sparse observations and deep learning.Frontiers in Marine Science 11 (2024), 1151868

    2024

  2. [10]

    Grace Deng, Cuize Han, and David S Matteson. 2022. Extended missing data imputation via GANs for ranking applications. Data Mining and Knowledge Discovery 36, 4 (2022), 1498–1520

    2022

  3. [11]

    Weiwei Fu, Jun She, and Shiyu Zhuang. 2011. Application of an Ensemble Optimal Interpolation in a North/Baltic Sea model: Assimilating temperature and salinity profiles. Ocean Modelling 40, 3-4 (2011), 227–245

    2011

  4. [12]

    Ian J Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative adversarial nets. Advances in neural information processing systems 27 (2014)

    2014

  5. [13]

    Nicolas Guillou, Georges Chapalain, and Sébastien Petton. 2023. Predicting sea surface salinity in a tidal estuary with machine learning. Oceanologia 65, 2 (2023), 318–332

    2023

  6. [14]

    Shengnan Guo, Tonglong Wei, Yiheng Huang, Miaomiao Zhao, Ran Chen, Yan Lin, Youfang Lin, and Huaiyu Wan. 2024. An Experimental Evaluation of Imputation Models for Spatial-Temporal Traffic Data. arXiv:2412.04733 [cs.LG] https://arxiv. org/abs/2412.04733

    2024

  7. [15]

    Michael Hahn. 2020. Theoretical limitations of self-attention in neural sequence models. Transactions of the Association for Computational Linguistics 8 (2020), 156–171

    2020

  8. [16]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation 9, 8 (1997), 1735–1780

    1997

  9. [17]

    Verena Hormann, Luca R Centurioni, and Gilles Reverdin. 2015. Evaluation of drifter salinities in the subtropical North Atlantic. Journal of Atmospheric and Oceanic Technology 32, 1 (2015), 185–192

    2015

  10. [18]

    Uiwon Hwang, Dahuin Jung, and Sungroh Yoon. 2019. Hexagan: Generative ad- versarial nets for real world classification. In International conference on machine learning. PMLR, 2921–2930

    2019

  11. [19]

    Xiaoyan Jia, Qiyan Ji, Lei Han, Yu Liu, Guoqing Han, and Xiayan Lin. 2022. Prediction of sea surface temperature in the East China Sea based on LSTM neural network. Remote Sensing 14, 14 (2022), 3300

    2022

  12. [20]

    Nan Jiang, Yanan Li, Hua Zuo, Hui Zheng, and Qinghe Zheng. 2020. BiLSTM-A: A missing value imputation method for PM2. 5 prediction. In 2020 2nd International Conference on Applied Machine Learning (ICAML) . IEEE, 23–28

    2020

  13. [21]

    Bora Jin, Amy H Herring, and David Dunson. 2024. Spatial predictions on physically constrained domains: applications to Arctic sea salinity data. The Annals of Applied Statistics 18, 2 (2024), 1596–1617

    2024

  14. [22]

    Ming Jin, Guangsi Shi, Yuan-Fang Li, Qingsong Wen, Bo Xiong, Tian Zhou, and Shirui Pan. 2023. How expressive are spectral-temporal graph neural networks for time series forecasting? arXiv preprint arXiv:2305.06587 (2023)

    2023 arXiv

  15. [23]

    Taesung Kim, Jinhee Kim, Yunwon Tae, Cheonbok Park, Jang-Ho Choi, and Jaegul Choo. 2022. Reversible Instance Normalization for Accurate Time-Series Forecasting against Distribution Shift. In International Conference on Learning Representations. https://openreview.net/forum?id=...

    2022

  16. [24]

    Yeo-Jin Kim and Min Chi. 2018. Temporal Belief Memory: Imputing Missing Data during RNN Training.. In In Proceedings of the 27th International Joint Conference on Artificial Intelligence (IJCAI-2018)

    2018

  17. [25]

    Young Jun Kim, Daehyeon Han, Eunna Jang, Jungho Im, and Taejun Sung. 2023. Remote sensing of sea surface salinity: challenges and research directions. GI- Science & Remote Sensing 60, 1 (2023), 2166377

    2023

  18. [26]

    Tuyen Pham Le, Cheolkyun Rho, Yelin Min, Sungreong Lee, and Daewoo Choi

  19. [27]

    Jiyue Li, Senzhang Wang, Jiaqiang Zhang, Hao Miao, Junbo Zhang, and Philip S Yu

  20. [28]

    Xiao Li, Huan Li, Harry Kai-Ho Chan, Hua Lu, and Christian S Jensen. 2023. Data imputation for sparse radio maps in indoor positioning. In 2023 IEEE 39th International Conference on Data Engineering (ICDE) . IEEE, 2235–2248

    2023

  21. [29]

    Meiling Liu, Xiangnan Liu, Jiale Jiang, and Xiaopeng Xia. 2013. Artificial neural network and random forest approaches for modeling of sea surface salinity. International Journal of Remote Sensing Applications 3, 4 (2013), 229–235

    2013

  22. [30]

    Yixin Liu, Thalaiyasingam Ajanthan, Hisham Husain, and Vu Nguyen. 2024. Self-supervision improves diffusion models for tabular data imputation. In Pro- ceedings of the 33rd ACM International Conference on Information and Knowledge Management. 1513–1522

    2024

  23. [31]

    Trevor J McDougall, Paul M Barker, Ryan M Holmes, Rich Pawlowicz, Stephen M Griffies, and Paul J Durack. 2021. The interpretation of temperature and salinity variables in numerical ocean model output and the calculation of heat fluxes and heat content. Geoscientific Model Deve...

    2021

  24. [32]

    Oleg Melnichenko, Peter Hacker, Nikolai Maximenko, Gary Lagerloef, and James Potemra. 2014. Spatial optimal interpolation of Aquarius sea surface salinity: Algorithms and implementation in the North Atlantic. Journal of Atmospheric and Oceanic Technology 31, 7 (2014), 1583–1600

    2014

  25. [33]

    Oleg Melnichenko, Peter Hacker, Nikolai Maximenko, Gary Lagerloef, and James Potemra. 2016. Optimum interpolation analysis of A quarius sea surface salinity. Journal of Geophysical Research: Oceans 121, 1 (2016), 602–616

    2016

  26. [34]

    Till Röthig, Stacey M Trevathan-Tackett, Christian R Voolstra, Cliff Ross, Samuel Chaffron, Paul J Durack, Laura M Warmuth, and Michael Sweet. 2023. Human- induced salinity changes impact marine organisms and ecosystems.Global change biology 29, 17 (2023), 4731–4749

    2023

  27. [35]

    Matthew W Schmidt, Howard J Spero, and David W Lea. 2004. Links between salinity variation in the Caribbean and North Atlantic thermohaline circulation. Nature 428, 6979 (2004), 160–163

    2004

  28. [36]

    Xu Shen, Yixin Liu, Yiwei Dai, Yili Wang, Rui Miao, Yue Tan, Shirui Pan, and Xin Wang. 2025. Understanding the Information Propagation Effects of Com- munication Topologies in LLM-based Multi-Agent Systems. arXiv preprint arXiv:2505.23352 (2025)

    2025 arXiv

  29. [37]

    Katie Smyth and Mike Elliott. 2016. Effects of changing salinity on the ecology of the marine environment. Stressors in the marine environment: physiological and ecological responses; societal implications (2016), 161–174

    2016

  30. [38]

    Tao Song, Zihe Wang, Pengfei Xie, Nisheng Han, Jingyu Jiang, and Danya Xu

  31. [39]

    Lizhuang Tan, Wei Su, Wei Zhang, Huiling Shi, Jingying Miao, and Pilar Manzanares-Lopez. 2021. A packet loss monitoring system for in-band network telemetry: detection, localization, diagnosis and recovery. IEEE Transactions on Network and Service Management 18, 4 (2021), 4151–4168

    2021

  32. [40]

    Olga Troyanskaya, Michael Cantor, Gavin Sherlock, Pat Brown, Trevor Hastie, Robert Tibshirani, David Botstein, and Russ B Altman. 2001. Missing value estimation methods for DNA microarrays. Bioinformatics 17, 6 (2001), 520–525

    2001

  33. [41]

    George Veronis. 1972. On properties of seawater defined by temperature, salinity, and pressure. (1972)

    1972

  34. [42]

    Elena Voita, David Talbot, Fedor Moiseev, Rico Sennrich, and Ivan Titov. 2019. Analyzing multi-head self-attention: Specialized heads do the heavy lifting, the rest can be pruned. arXiv preprint arXiv:1905.09418 (2019)

    2019 arXiv

  35. [43]

    Yufeng Wang, Dan Li, Xiang Li, and Min Yang. 2021. PC-GAIN: Pseudo-label conditional generative adversarial imputation networks for incomplete data. Neural Networks 141 (2021), 395–403

    2021

  36. [44]

    FA Whitney, WR Crawford, and PJ Harrison. 2005. Physical processes that enhance nutrient transport and primary productivity in the coastal and open ocean of the subarctic NE Pacific. Deep Sea Research Part II: Topical Studies in Oceanography 52, 5-6 (2005), 681–706

    2005

  37. [45]

    Jinsung Yoon, William R Zame, and Mihaela Van Der Schaar. 2018. Deep sensing: Active sensing using multi-directional recurrent neural networks. InInternational OASIS: Harnessing Diffusion Adversarial Network for Ocean Salinity Imputation using Sparse Drifter Trajectories CIKM ...

    2018

  38. [46]

    Seongwook Yoon and Sanghoon Sull. 2020. GAMIN: Generative adversarial mul- tiple imputation network for highly missing data. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 8456–8464

    2020

  39. [47]

    Lisan Yu. 2011. A global relationship between the ocean water cycle and near- surface salinity. Journal of Geophysical Research: Oceans 116, C10 (2011)

    2011

  40. [48]

    Xinlu Zhang, Shiyang Li, Zhiyu Chen, Xifeng Yan, and Linda Ruth Petzold. 2023. Improving medical predictions by irregular multimodal electronic health records modeling. In International Conference on Machine Learning . PMLR, 41300–41313

    2023

  41. [49]

    He Zhao, Ke Sun, Amir Dezfouli, and Edwin V Bonilla. 2023. Transformed distribution matching for missing value imputation. In International Conference on Machine Learning. PMLR, 42159–42186

    2023

  42. [50]

    Xin Zheng, Wei Huang, Chuan Zhou, Ming Li, and Shirui Pan. 2025. Test- Time Graph Neural Dataset Search With Generative Projection. In Forty-second International Conference on Machine Learning

    2025

  43. [51]

    Xin Zheng, Dongjin Song, Qingsong Wen, Bo Du, and Shirui Pan. 2024. On- line GNN Evaluation Under Test-time Graph Distribution Shifts. In The Twelfth International Conference on Learning Representations

    2024

  44. [52]

    Xin Zheng, Miao Zhang, Chunyang Chen, Soheila Molaei, Chuan Zhou, and Shirui Pan. 2024. Gnnevaluator: Evaluating gnn performance on unseen graphs without labels. Advances in Neural Information Processing Systems 36 (2024)

    2024

  45. [53]

    Xin Zheng, Miao Zhang, Chunyang Chen, Quoc Viet Hung Nguyen, Xingquan Zhu, and Shirui Pan. 2024. Structure-free graph condensation: From large-scale graphs to condensed graph-free data. Advances in Neural Information Processing Systems 36 (2024)

    2024

  46. [54]

    Xu Zhou, Xiaofeng Liu, Gongjin Lan, and Jian Wu. 2021. Federated condi- tional generative adversarial nets imputation method for air quality missing data. Knowledge-Based Systems 228 (2021), 107261

    2021

  47. [2020]

    Journal of Atmospheric and Oceanic Technology 37, 2 (2020), 317–325

    A novel dual path gated recurrent unit model for sea surface salinity prediction. Journal of Atmospheric and Oceanic Technology 37, 2 (2020), 317–325

    2020

  48. [2021]

    IEEE Access 9 (2021), 137165–137175

    A2GAN: A deep reinforcement-based learning algorithm for risk-aware in finance. IEEE Access 9 (2021), 137165–137175

    2021

  49. [2022]

    IEEE Transactions on Knowledge and Data Engineering 35, 6 (2022), 5851–5864

    Fine-grained urban flow inference with incomplete data. IEEE Transactions on Knowledge and Data Engineering 35, 6 (2022), 5851–5864

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.