Pith. sign in

REVIEW 3 major objections 6 minor 52 references

Text embedding models can be great data engineers

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

Pith's one-line read Frozen text embeddings can replace most of the data-engineering pipeline for time-series classification.

desk verdict New combination, shaky attribution: the v2.0 gains likely come from the supervised VIB, not the text embeddings, but the idea is novel enough to merit a serious revision. read the letter →

arxiv 2505.14802 v1 pith:HAR74GUX submitted 2025-05-20 cs.LG

classification cs.LG MSC 68T0768T0562M10
keywords timeseriesclassificationtextembeddingsvariationalinformationbottleneckdataengineeringautomationmultivariaterepresentationlearningrawformatfrozen
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes ADEPT, a classification pipeline that skips the conventional data-engineering chain of imputation, normalization, feature extraction, and feature selection. Instead, it serializes raw multivariate time series into text, feeds that text through a frozen general-purpose text-embedding model, and classifies the resulting vectors. The central claim is that these text embeddings carry as much task-relevant temporal and cross-channel structure as hand-engineered numeric features, so the expensive domain-specific parts of a pipeline become unnecessary. On four tasks, ADEPT v2.0 reaches 97.83%, 73.68%, 88.49%, and 74.35% accuracy, outperforming the cited application-specific baselines on three of four tasks and approaching the best on the fourth. If the claim holds, an off-the-shelf embedder plus a small variational bottleneck can replace weeks of domain-specific data engineering for time-series classification.

What carries the argument

The machinery has three parts. First, the raw-format representation (RFR): each temporal view is split into M chunks, each chunk is serialized into a text string, and a frozen text-embedding function maps it to a vector. Second, a per-view variational information bottleneck (VIB) encoder parameterizes a Gaussian latent distribution over the embedding, uses the reparameterization trick to sample codes, and trains with cross-entropy plus a KL-divergence penalty to a standard normal prior; this is the component that compresses away noise and variance. Third, a transformer autoencoder is pretrained to reconstruct the fused sequences and then fine-tuned to classify, with per-view probabilities combined by a product-of-experts fusion. The entropy-equivalence claim, that text embeddings of serialized raw series carry as much information as engineered numeric features, is the stated reason the first component can replace the whole data-engineering chain.

What would settle it

A concrete test: on each dataset, rerun ADEPT v2.0 after replacing the serialized numeric values with random numbers in the same format, keeping labels, chunking, VIB, and transformer training unchanged. If accuracy stays near the reported levels, the text embedder's reading of the actual numbers is not what carries the signal; if accuracy collapses to chance, the text-embedding transfer claim is supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that the entropy of embeddings of a raw serialized time series is equivalent, or in many cases superior, to that of numerically dense feature vectors produced by a conventional data-engineering pipeline; therefore a frozen general-purpose text embedder can serve as the representation layer for raw multivariate time-series classification. ADEPT v2.0 realizes this by chunking each series, serializing each chunk into a character string of timestamps and channel readings, embedding that string with a frozen text-embedding model, compressing the embeddings through a variational information bottleneck trained per view with a classification loss plus a KL penalty, and classifying the fused codes with a transformer. In the paper's experiments, the pipeline reaches 97.83% on PLAsTiCC light curves, 73.68% on SelfRegulationSCP2 EEG, 88.49% on Bitcoin 2023 next-day movement, and 74.35% on the HRI hydropower fault task. The VIB step is load-bearing: ADEPT v1.0 without it scores 95.98%, 58.97%, 45.40%, and 45.00% on the same tasks, so the raw text embeddings alone are not enough.

Load-bearing premise

The load-bearing premise is that serializing numbers into text and embedding that text preserves the temporal and cross-channel information that hand-engineered features are built to capture; the paper asserts this equivalence but never measures it.

Editorial extensions

If this is right

  • Conventional steps such as imputation, normalization, outlier removal, and feature selection can be skipped for many multivariate time-series classification tasks without sacrificing accuracy.
  • Because categorical and numeric channels are embedded in one serialized string, there is no need to design separate encodings for mixed-type sensor and event data.
  • The large gap between ADEPT v1.0 and v2.0 implies that a supervised variational bottleneck is required to turn noisy frozen embeddings into usable representations.
  • The same frozen-embedding-plus-bottleneck recipe can be dropped into other classifier families such as MLPs or SVMs, and can run with on-premise embedders where data cannot leave the facility.
  • End-to-end automation of the representation step removes the weeks-long manual iteration cycle typical of pipeline design for new domains.

Reading between the lines

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

  • Inference: The entropy-equivalence premise could be made quantitative by measuring mutual information between serialized-text embeddings and the original numeric series; if that number is low, the performance would have to be credited to the label-trained VIB rather than to the text embedder.
  • Inference: The same recipe should be testable on regression and forecasting, not just classification; the transfer claim, if true, is about representation rather than about the classification head.
  • Inference: The very low v1.0 accuracy on Bitcoin suggests that raw text embeddings can pick up surface formatting regularities; a controlled ablation with shuffled or masked numbers would reveal whether the embedder is exploiting numeric content or textual patterns.
  • Inference: The HRI top-2 accuracy of 97.5% hints that the bottleneck preserves enough structure for ranked decision support, which could matter in maintenance settings; calibrating these top-k confidences for human-in-the-loop use is a natural extension.
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. The paper proposes ADEPT, a pipeline that serializes raw multivariate time series into text, embeds the chunks with a frozen pretrained text embedding model, optionally compresses the embeddings with a variational information bottleneck (VIB), and then classifies with a transformer. Two versions are compared: v1.0 (text embeddings plus transformer) and v2.0 (adding VIB). Experiments cover astronomy (PLAsTiCC), EEG (SelfRegulationSCP2), Bitcoin price trend, and hydropower fault data. The paper reports that v2.0 achieves 97.83%, 73.68%, 88.49%, and 74.35% accuracy on these tasks, beating cited literature baselines on three of four, and claims that general-purpose text embeddings can replace data engineering steps for time series classification.

Significance. If the claim were established, the paper would be a useful contribution: it would show that frozen text embedding models, originally trained for natural language, can serve as generic representation backbones for heterogeneous time series, potentially reducing manual feature engineering in several applied domains. The paper also explores a sensible design idea, namely using a VIB layer to denoise high-dimensional text embeddings. The diversity of the four evaluated datasets is a strength, and the paper is generally clearly written. However, the current evidence does not isolate the contribution of the text embedding model itself. The v1.0 results are near or below competitive baselines, while v2.0's gains coincide with the addition of a label-trained VIB. Without a control that replaces the pretrained embedder with a random or non-semantic encoder, the central assertion that text embeddings are 'great data engineers' is not yet supported. The paper also compares against published accuracies from other papers rather than a shared, rerun experimental protocol, which weakens the quantitative claims.

major comments (3)
  1. [Section 5, Table 1, Eq. (8)] The central claim that frozen text embeddings, without feature engineering, are powerful raw time-series representations is not supported by the ablation shown in Table 1. ADEPT v1.0, which uses text embeddings directly, scores 58.97% on SelfRegulationSCP2, 45.40% on Bitcoin, and 45.00% on HRI, while ADEPT v2.0, which adds a VIB encoder trained with the label-dependent loss in Eq. (8), jumps to 73.68%, 88.49%, and 74.35%. Because the VIB encoder is trained with the classification loss, the performance gains could be due to the supervised learned encoder rather than to any semantic or structural prior in pretrained text embeddings. A necessary control is to replace the frozen text embedder with a random projection or a randomly initialized fixed encoder, keeping the VIB and transformer identical, and show that ADEPT v2.0 with pretrained text embeddings still outperforms that control. Without such an ablation, the paper's title-level claim that text embedding models are great data engineers is not established.
  2. [Section 1 and Abstract] The load-bearing premise of the paper, stated in Section 1, is that 'the entropy of embeddings corresponding to textually dense raw format representation of timeseries can be intuitively viewed as equivalent (or in many cases superior) to that of numerically dense vector representations obtained by data engineering pipelines.' This equivalence is never defined, measured, or independently demonstrated. The VIB step explicitly compresses the text embeddings and is trained with labels, so any entropy reduction or task-relevant structure may come from the VIB rather than from the text embedding model. The paper should either provide an operational definition and measurement of the claimed entropy equivalence or substantially weaken the causal interpretation and present ADEPT v2.0 as a full learned pipeline rather than as evidence that text embeddings alone can leapfrog data engineering.
  3. [Tables 1a–1d, Section 4] The benchmark comparisons are not apples-to-apples. The accuracies for CATS, AMPEL, ORACLE, MiniRocket, MHCAN, TSEM, RLSTM, EDL, and DQN are taken from their respective papers, which use different datasets, preprocessing choices, and train/test splits. For example, the PLAsTiCC results in [10, 33, 45] are reported on their own evaluation setups, and the paper does not show that the same 7,848 light curves and the same split are used. The claim of '>10%' improvement over the best benchmark on PLAsTiCC is therefore not a controlled comparison. The authors should either rerun the baselines on the exact same train/validation/test splits used for ADEPT, or clearly present the comparison as a cross-paper reference and avoid direct percentage-improvement claims.
minor comments (6)
  1. [Appendix D versus Section 3] Appendix D repeatedly mentions 'clustering pretraining' for the transformer classifier, but the methodology in Section 3.4 describes only an autoencoder pretraining stage (Eq. 11) and a classification fine-tuning stage (Eq. 12); the clustering objective is never defined. Please add the missing definition or remove the term.
  2. [Algorithm 2, Eq. (11)] Algorithm 2's autoencoder loss is written with a normalization factor of 1/(3N) and a denominator of M^2E, while Eq. (11) uses 1/(K·N) and 2ME. These are inconsistent; please align the notation and the constants.
  3. [References [4] and [5]] References [4] and [5] appear to be duplicate entries for the same TSFEL library, with different author lists and page numbers. Please merge them into a single correct citation.
  4. [Appendix B.2, Table 1b] Appendix B.2 states that SelfRegulationSCP2 has eight scalp channels, but Table 1b and the main text describe the input as 7-channel EEG. Please reconcile this discrepancy.
  5. [Appendix B.4, Table 6] Appendix B.4 says the HRI dataset has 390 events split chronologically into 80%/10%/10%, which would give 39 test events, but Table 6 lists 41 event IDs. Please clarify the actual test set size.
  6. [General] The paper states it provides an 'end-to-end automated implementation' but does not include code, data, or a reproducibility statement. Given the heavy reliance on proprietary or API-based embedding models, a reproducibility appendix with exact prompts, serialization formats, and random seeds would be valuable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark study; its headline claim is supported by held-out test accuracy, not by construction, and no load-bearing step reduces to its own inputs or to self-citations.

full rationale

The paper's central assertion, that general-purpose text embeddings can serve as powerful raw time-series representations, is an empirical claim evaluated on held-out test splits (e.g., 97.83% on PLAsTiCC, 73.68% on SelfRegulationSCP2, 88.49% on Bitcoin 2023, 74.35% on HRI). The VIB encoder is trained with labels through Eq. 8, but that is ordinary supervised training followed by test-set evaluation, not a fitted input renamed as a prediction. The 'entropy equivalence' stated in Section 1 is a motivating analogy, not a formal premise: no equation in the methodology derives accuracy from it, and no first-principles result is claimed from it. There are no load-bearing self-citations: the cited works are external baselines and standard references, and the authors do not invoke a prior 'uniqueness theorem' or ansatz from their own earlier work to force the framework. The absence of a random-embedding control is a legitimate attribution/ablation concern, and the entropy-equivalence statement is unmeasured, but neither constitutes circularity under the required definition: the empirical results are not equivalent to the inputs by construction. The paper's contributions are testable and externally benchmarked, so the correct circularity finding is none.

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

The framework introduces no new physical or mathematical entities. It rests on transfer assumptions about text embeddings, on an unmeasured entropy-equivalence assertion, and on the comparability of previously published baseline accuracies. The many per-dataset hyperparameters are fitted choices, not derived quantities.

free parameters (5)
  • chunk count M per dataset = 10 (PLAsTiCC), 24 (SelfRegulationSCP2), 5 (Bitcoin), 24 (HRI)
    Chunk length L=T/M controls the granularity of text serialization; authors say M can be chosen by domain insight or validation studies, but report no sensitivity analysis. Section 3.1, Appendix D.
  • VIB bottleneck dimension d = 256
    Set identically for all datasets; no ablation. Appendix D.
  • VIB trade-off weight beta = 1e-4
    Set identically for all datasets; no sensitivity analysis. Appendix D.
  • Transformer hyperparameters = h=128, nhead=32 or 16, layers=2, dimff=128
    Chosen per dataset without ablations. Appendix D.
  • number of pretraining and classification epochs = EVIB=100, EAE=50/100, ECL=50/200
    Different across datasets; no convergence or variance analysis. Appendix D.
assumptions (4)
  • ad hoc to paper Frozen text embedding models trained on natural language preserve the temporal and cross-channel structure of serialized numeric time series.
    This is the load-bearing transfer assumption in Section 1 and Section 3.2; no evidence is provided beyond final accuracies.
  • ad hoc to paper The entropy of text embeddings of raw text representations is equivalent or superior to the entropy of engineered numeric feature vectors for classification.
    Intro Section 1 asserts this 'intuitively', but it is never defined, measured, or proven.
  • domain assumption Accuracy values reported in prior papers for CATS, AMPEL, ORACLE, MiniRocket, MHCAN, TSEM, RLSTM, EDL, and DQN can be compared to ADEPT numbers without rerunning those methods under identical splits.
    Section 4 compares all results directly in Tables 1a-1d; the paper never re-executes baselines.
  • domain assumption The OpenAI and nomic embedding APIs produce stable, reproducible embeddings over time.
    Implementation in Appendix D relies on hosted APIs; model versions and access dates are not pinned.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Text embedding models can be great data engineers." pith.science (2026). https://pith.science/paper/HAR74GUX

@misc{pith2026250514802,
  author       = {Pith},
  title        = {Pith review of: Text embedding models can be great data engineers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HAR74GUX}},
  note         = {Machine review of arXiv:2505.14802}
}
read the original abstract

Data engineering pipelines are essential - albeit costly - components of predictive analytics frameworks requiring significant engineering time and domain expertise for carrying out tasks such as data ingestion, preprocessing, feature extraction, and feature engineering. In this paper, we propose ADEPT, an automated data engineering pipeline via text embeddings. At the core of the ADEPT framework is a simple yet powerful idea that the entropy of embeddings corresponding to textually dense raw format representation of time series can be intuitively viewed as equivalent (or in many cases superior) to that of numerically dense vector representations obtained by data engineering pipelines. Consequently, ADEPT uses a two step approach that (i) leverages text embeddings to represent the diverse data sources, and (ii) constructs a variational information bottleneck criteria to mitigate entropy variance in text embeddings of time series data. ADEPT provides an end-to-end automated implementation of predictive models that offers superior predictive performance despite issues such as missing data, ill-formed records, improper or corrupted data formats and irregular timestamps. Through exhaustive experiments, we show that the ADEPT outperforms the best existing benchmarks in a diverse set of datasets from large-scale applications across healthcare, finance, science and industrial internet of things. Our results show that ADEPT can potentially leapfrog many conventional data pipeline steps thereby paving the way for efficient and scalable automation pathways for diverse data science applications.

Figures

Figures reproduced from arXiv: 2505.14802 by the authors.

Figure 1
Figure 1. Comparison of the model and benchmark. The methodological contri￾bution of ADEPT relies on exploiting text embedding models primarily geared for LLM use cases as the foun￾dational building block to power timeseries oriented predictive analytics tasks. As a result, the ADEPT framework leverages LLM based text embedding mod￾els as a black box, eliminat￾ing need for complex fine-tuning and retraining tasks. We introduc… view at source ↗
Figure 2
Figure 2. Illustration of the ADEPT v2.0. Framework [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. t-SNE projection of segment embeddings (ADEPT v2.0) across different applications. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: 3D t-SNE projection of 1536-dim segment embeddings from the PLAsTiCC-2018 LSST [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Normalized confusion matrix for the IB-filtered pipeline on the PLAsTiCC-2018 LSST [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: 3D t-SNE projection of 1536-dim segment embeddings for SelfRegulationSCP2 dataset, [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Normalized confusion matrix for the ADEPT v2.0 pipeline on the SelfRegulationSCP2 [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: 2D t-SNE projection of text-serialized Bitcoin segment embeddings (1536 D), colored by [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Normalized confusion matrix for the IB-filtered pipeline on the Bitcoin market dataset. [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: 3D t-SNE projection of 15 min segment embeddings from the HRI dataset, colored by failure cause code (13 classes). Left: raw 768-dim Nomic text embeddings; Right: embeddings after IB filtering [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: Normalized confusion matrix for the IB-filtered pipeline on HRI. Rows correspond to true [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 45 canonical work pages

  1. [1]

    Information dropout: Learning optimal representations through noisy computation

    Alessandro Achille and Stefano Soatto. Information dropout: Learning optimal representations through noisy computation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(12):2897–2905, 2018

  2. [2]

    Deep variational information bottleneck

    Alexander A Alemi, Ian Fischer, Joshua V Dillon, and Kevin Murphy. Deep variational information bottleneck. arXiv preprint arXiv:1612.00410, 2016

  3. [3]

    A comprehensive review on machine learning in healthcare industry: classification, restrictions, opportunities and challenges.Sensors, 23(9):4178, 2023

    Qi An, Saifur Rahman, Jingwen Zhou, and James Jin Kang. A comprehensive review on machine learning in healthcare industry: classification, restrictions, opportunities and challenges.Sensors, 23(9):4178, 2023

  4. [4]

    TSFEL: Time series feature extraction library

    Marília Barandas, Duarte Folgado, Letícia Fernandes, Sara Santos, Mariana Abreu, Patrícia Bota, Hui Liu, Tanja Schultz, and Hugo Gamboa. TSFEL: Time series feature extraction library. SoftwareX, 11:100456, 2020

  5. [5]

    TSFEL: Time series feature extraction library

    Pablo Barandas, Víctor de la Iglesia, et al. TSFEL: Time series feature extraction library. SoftwareX, 11:100453, 2020

  6. [6]

    Financial time series forecasting-a deep learning approach

    Alexiei Dingli and Karl Sant Fournier. Financial time series forecasting-a deep learning approach. International Journal of Machine Learning and Computing, 7(5):118–122, 2017

  7. [7]

    Information transmission with additional noise

    R Dobrushin and Boris Tsybakov. Information transmission with additional noise. IRE Transactions on Information Theory, 8(5):293–304, 1962

  8. [8]

    Systematic literature review of preprocessing techniques for imbalanced data

    Ebubeogu Amarachukwu Felix and Sai Peck Lee. Systematic literature review of preprocessing techniques for imbalanced data. Iet Software, 13(6):479–496, 2019

Show all 52 references
  1. [9]

    Springenberg, Manuel Blum, and Frank Hutter

    Matthias Feurer, Aaron Klein, Katharina Eggensperger, Jost H. Springenberg, Manuel Blum, and Frank Hutter. Efficient and robust automated machine learning. In Advances in Neural Information Processing Systems, volume 28, 2015

  2. [10]

    Transient classifiers for fink-benchmarks for lsst

    BMO Fraga, CR Bom, A Santos, E Russeil, M Leoni, J Peloton, EEO Ishida, A Möller, and S Blondin. Transient classifiers for fink-benchmarks for lsst. Astronomy & Astrophysics, 692:A208, 2024

  3. [11]

    The information bottleneck problem and its applications in machine learning

    Ziv Goldfeld and Yury Polyanskiy. The information bottleneck problem and its applications in machine learning. IEEE Journal on Selected Areas in Information Theory, 1(1):19–38, 2020

  4. [12]

    Audioclip: Extending clip to image, text and audio

    Andrey Guzhov, Federico Raue, Jörn Hees, and Andreas Dengel. Audioclip: Extending clip to image, text and audio. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 976–980. IEEE, 2022

  5. [13]

    Distributional structure

    Zellig S Harris. Distributional structure. Word, 10(2-3):146–162, 1954

  6. [14]

    beta-vae: Learning basic visual concepts with a constrained variational framework

    Irina Higgins, Loic Matthey, Arka Pal, Christopher Burgess, Xavier Glorot, Matthew Botvinick, Shakir Mohamed, and Alexander Lerchner. beta-vae: Learning basic visual concepts with a constrained variational framework. In International Conference on Learning Representations, 2017

  7. [15]

    A survey of outlier detection methodologies

    Victoria Hodge and Jim Austin. A survey of outlier detection methodologies. Artificial Intelligence Review, 22:85–126, 2004

  8. [16]

    A framework for extracting urban functional regions based on multiprototype word embeddings using points-of- interest data

    Sheng Hu, Zhanjun He, Liang Wu, Li Yin, Yongyang Xu, and Haifu Cui. A framework for extracting urban functional regions based on multiprototype word embeddings using points-of- interest data. Computers, Environment and Urban Systems, 80:101442, 2020

  9. [17]

    Multilevel temporal-spectral fusion network for multivariate time series classification

    Xulin Huang, Shizhe Ding, Xinru Zhang, Jingyan Sui, Yue Yu, and Bu Dongbo. Multilevel temporal-spectral fusion network for multivariate time series classification. In2024 International Joint Conference on Neural Networks (IJCNN), pages 1–7. IEEE, 2024

  10. [18]

    HRI Dataset

    Hydropower Research Institute. HRI Dataset. https://www.hridata.org, 2025. 10

  11. [19]

    Large-scale representation learning from visually grounded untranscribed speech

    Gabriel Ilharco, Yuan Zhang, and Jason Baldridge. Large-scale representation learning from visually grounded untranscribed speech. arXiv preprint arXiv:1909.08782, 2019

  12. [20]

    Autokeras: An automl library for deep learning

    Haifeng Jin, François Chollet, Qingquan Song, and Xia Hu. Autokeras: An automl library for deep learning. Journal of Machine Learning Research, 24(6):1–6, 2023

  13. [21]

    Dimensionality reduction for fast similarity search in large time series databases

    Eamonn Keogh, Kaushik Chakrabarti, Michael Pazzani, and Sharad Mehrotra. Dimensionality reduction for fast similarity search in large time series databases. Knowledge and Information Systems, 3:263–286, 2001

  14. [22]

    Representation learning of clinical multivariate time series with random filter banks

    Alireza Keshavarzian, Hojjat Salehinejad, and Shahrokh Valaee. Representation learning of clinical multivariate time series with random filter banks. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023

  15. [23]

    Time series classification of cryptocurrency price trend based on a recurrent lstm neural network

    Do-Hyung Kwon, Ju-Bong Kim, Ju-Sung Heo, Chan-Myung Kim, and Youn-Hee Han. Time series classification of cryptocurrency price trend based on a recurrent lstm neural network. Journal of Information Processing Systems, 15(3):694–706, 2019

  16. [24]

    H2o automl: Scalable automatic machine learning

    Erin LeDell and Said Poirier. H2o automl: Scalable automatic machine learning. In AutoML Workshop at the 37th International Conference on Machine Learning, 2020

  17. [25]

    Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median

    Christophe Leys, Christophe Ley, Olivier Klein, Philippe Bernard, and Laurent Licata. Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 49(4):764–766, 2013

  18. [26]

    Feature selection: A data perspective

    Jundong Li, Kewei Cheng, Suhang Wang, Fred Morstatter, Robert P Trevino, Jiliang Tang, and Huan Liu. Feature selection: A data perspective. ACM Computing Surveys (CSUR), 50(6):1–45, 2017

  19. [27]

    Missing value imputation: a review and analysis of the literature (2006–2017)

    Wei-Chao Lin and Chih-Fong Tsai. Missing value imputation: a review and analysis of the literature (2006–2017). Artificial Intelligence Review, 53:1487–1509, 2020

  20. [28]

    Cubic spline interpolation

    Sky McKinley and Megan Levine. Cubic spline interpolation. College of the Redwoods , 45(1):1049–1060, 1998

  21. [29]

    Howto100m: Learning a text-video embedding by watching hundred million narrated video clips

    Antoine Miech, Dimitri Zhukov, Jean-Baptiste Alayrac, Makarand Tapaswi, Ivan Laptev, and Josef Sivic. Howto100m: Learning a text-video embedding by watching hundred million narrated video clips. In Proceedings of the IEEE/CVF international conference on computer vision, pages ...

  22. [30]

    Enhanced bitcoin price direction forecasting with dqn

    Azamjon Muminov, Otabek Sattarov, and Daeyoung Na. Enhanced bitcoin price direction forecasting with dqn. IEEE Access, 12:29093–29112, 2024

  23. [31]

    Unsupervised embedding of trajectories captures the latent structure of scientific migration

    Dakota Murray, Jisung Yoon, Sadamori Kojaku, Rodrigo Costas, Woo-Sung Jung, Staša Miloje- vi´c, and Yong-Yeol Ahn. Unsupervised embedding of trajectories captures the latent structure of scientific migration. Proceedings of the National Academy of Sciences, 120(52):e2305414120, 2023

  24. [32]

    Delineating urban functional use from points of interest data with neural network embedding: A case study in greater london

    Haifeng Niu and Elisabete A Silva. Delineating urban functional use from points of interest data with neural network embedding: A case study in greater london. Computers, Environment and Urban Systems, 88:101651, 2021

  25. [33]

    Ampel workflows for lsst: Modular and reproducible real-time photometric classification

    Jakob Nordin, Valery Brinnel, Jakob van Santen, Simeon Reusch, and Marek Kowalski. Ampel workflows for lsst: Modular and reproducible real-time photometric classification. arXiv preprint arXiv:2501.16511, 2025

  26. [34]

    Nomic embed: Training a reproducible long context text embedder

    Zach Nussbaum, John X Morris, Brandon Duderstadt, and Andriy Mulyar. Nomic embed: Training a reproducible long context text embedder. arXiv preprint arXiv:2402.01613, 2024

  27. [35]

    Olson, Nathan Bartley, Ryan J

    Randal S. Olson, Nathan Bartley, Ryan J. Urbanowicz, and Jason H. Moore. Evaluation of a tree-based pipeline optimization tool for automating data science. In Proceedings of the Genetic and Evolutionary Computation Conference, pages 485–492, Denver, Colorado, USA, 2016. 11

  28. [36]

    text-embedding-3-small

    OpenAI. text-embedding-3-small. Computer software, Retrieved from https://platform.openai.com/docs/guides/embeddings, 2023. [Accessed May 15, 2025]

  29. [37]

    Tsem: Temporally-weighted spatiotemporal explainable neural network for multivariate time series

    Anh-Duy Pham, Anastassia Kuestenmacher, and Paul G Ploeger. Tsem: Temporally-weighted spatiotemporal explainable neural network for multivariate time series. In Future of Information and Communication Conference, pages 183–204. Springer, 2023

  30. [38]

    Convolutional neural net- work fault classification based on time-series analysis for benchmark wind turbine machine

    Reihane Rahimilarki, Zhiwei Gao, Nanlin Jin, and Aihua Zhang. Convolutional neural net- work fault classification based on time-series analysis for benchmark wind turbine machine. Renewable Energy, 185:916–931, 2022

  31. [39]

    Modelling data pipelines

    Aiswarya Raj, Jan Bosch, Helena Holmström Olsson, and Tian J Wang. Modelling data pipelines. In 2020 46th Euromicro conference on software engineering and advanced applications (SEAA), pages 13–20. IEEE, 2020

  32. [40]

    Time-series cryptocurrency forecasting using ensemble deep learning

    K Rama Rao, M Lakshmi Prasad, G Ravi Kumar, R Natchadalingam, Mohammad Manzoor Hussain, and Pundru Chandra Shaker Reddy. Time-series cryptocurrency forecasting using ensemble deep learning. In 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT), ...

  33. [41]

    Word embeddings for the analysis of ideological placement in parliamentary corpora

    Ludovic Rheault and Christopher Cochrane. Word embeddings for the analysis of ideological placement in parliamentary corpora. Political analysis, 28(1):112–133, 2020

  34. [42]

    Predicting high-level human judgment across diverse behavioral domains

    Russell Richie, Wanling Zou, and Sudeep Bhatia. Predicting high-level human judgment across diverse behavioral domains. Collabra: Psychology, 5(1):50, 2019

  35. [43]

    Gaussian processes for time-series modelling

    Stephen Roberts, Michael Osborne, Mark Ebden, Steven Reece, Neale Gibson, and Suzanne Aigrain. Gaussian processes for time-series modelling. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(1984):20110550, 2013

  36. [44]

    Automl: A systematic review on automated machine learning with neural architecture search

    Imrus Salehin, Md Shamiul Islam, Pritom Saha, SM Noman, Azra Tuni, Md Mehedi Hasan, and Md Abu Baten. Automl: A systematic review on automated machine learning with neural architecture search. Journal of Information and Intelligence, 2(1):52–81, 2024

  37. [45]

    Oracle: A real-time, hierarchical, deep-learning photometric classifier for the lsst

    Ved G Shah, Alex Gagliano, Konstantin Malanchev, Gautham Narayan, LSST Dark Energy Sci- ence Collaboration, et al. Oracle: A real-time, hierarchical, deep-learning photometric classifier for the lsst. arXiv preprint arXiv:2501.01496, 2025

  38. [46]

    Videobert: A joint model for video and language representation learning

    Chen Sun, Austin Myers, Carl V ondrick, Kevin Murphy, and Cordelia Schmid. Videobert: A joint model for video and language representation learning. In Proceedings of the IEEE/CVF international conference on computer vision, pages 7464–7473, 2019

  39. [47]

    Survey: Time-series data preprocessing: A survey and an empirical analysis

    Amal Tawakuli, Bastian Havers, Vincenzo Gulisano, Daniel Kaiser, and Thomas Engel. Survey: Time-series data preprocessing: A survey and an empirical analysis. Journal of Engineering Research, 2024

  40. [48]

    up”) or decrease (“down

    Peng Wang, Ke Wang, Yafei Song, and Xiaodan Wang. Autoldt: a lightweight spatio-temporal decoupling transformer framework with automated machine learning method for time series classification. Scientific Reports, 14:29801, 2024. 12 A ADEPT v2.0 Algorithm The full pseducode of ...

  41. [49]

    Feature extraction + Classifier: We linearly interpolate missing readings onto a uniform 30 s grid, slide 15 min windows over each event, extract over 9,000 time- and frequency-domain features per channel via TSFEL, select the top 100 via mutual information, normalize, and cla...

  42. [50]

    ADEPT v1.0: We serialize each 15 min segment and embed it offline to a 768-dim vector via the nomic-embed-text-v1 model, then classify directly

  43. [51]

    negativity

    ADEPT v2.0: Our full pipeline, where VIB compresses the 768 -dim embeddings before fusion and classification. E Detailed Results on Predicting the Astrophysical Class of Light Curves Figure 4 presents a 3D t-SNE projection of 1536-dimensional segment embeddings from the PLAsTi...

  44. [7009]

    Table 5: Per-class performance on the HRI dataset

    suffer from zero recall. Table 5: Per-class performance on the HRI dataset. Class Precision Recall F 1–score 3620 1.00 0.40 0.57 3710 0.50 0.50 0.50 4560 1.00 0.50 0.67 7009 0.00 0.00 0.00 7030 1.00 1.00 1.00 7050 1.00 1.00 1.00 7099 0.43 1.00 0.60 7110 0.89 0.76 0.82 9696 0.7...

Pith tools

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