Pith. sign in

REVIEW 3 major objections 6 minor 47 references

Feature Augmentations for High-Dimensional Learning

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Nonlinear factor augmentation improves stock-return prediction and portfolio returns across five learning algorithms.

desk verdict The paper's useful empirical claim—augmenting X with factors from nonlinear transforms often helps—is real, but the evaluation protocol (best-of-three, no error bars, supervised FNN in the mix) makes 'consistent improvement' hard to pin down. read the letter →

arxiv 2509.00232 v1 pith:WAIQ3KSL submitted 2025-08-29 stat.AP stat.ML

Xiaonan Zhu , Bingyan Wang , Jianqing Fan This is my paper
classification stat.APstat.ML MSC 62H2562J07
keywords featureaugmentationfactormodelsprincipalcomponentanalysisdiversifiedprojectionstockreturnpredictionChinesefinancialnewstextdatanonlineartransformations
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

High-dimensional features are usually correlated, so a few latent factors carry much of their signal. The paper's idea is to extract those factors not only from the raw design matrix but also from cheap nonlinear transformations of it—pairwise interactions, kernel matrices, and neural-network hidden layers—and then feed the factors together with the residualized original features into any supervised learner. Applied to predicting stock returns from roughly 914,000 Chinese financial news articles, the augmentation improves out-of-sample R2 for Lasso, Ridge, random forests, gradient boosted trees, and neural networks, and it sharpens event-study responses and long-short portfolio returns. The same recipe also improves classification and regression accuracy across datasets from image, text, biology, and finance. If correct, the claim matters because it sits between data collection and algorithm design: a simple PCA-style preprocessing step can buy prediction gains with almost no added computation.

What carries the argument

The load-bearing object is the augmented design (F, U): estimate K latent factors F from a transformed matrix (e.g., X_inter, the n × p(p+1)/2 matrix of pairwise products), then set U = X − F B-hat^⊤, the projection residual that removes the factor span from the original features. Because the span of (F, U) equals the span of (F, X), no information is discarded, while correlations among the original variables are weakened. Factors are estimated either by PCA or by diversified projection, in which a pretrained weight matrix W = √p(ξ-hat′_1,…,ξ-hat′_K′) built from n′ ≈ K² log p independent samples projects the high-dimensional data down to K dimensions before least-squares loading estimation;

What would settle it

Re-run the Chinese-news stock-return experiments with a pretraining sample of n′ ≈ 2,500 (or full PCA on a subsample) and compare out-of-sample R². If the augmentation gains shrink to zero or change sign, the reported improvements are an artifact of underpowered factor estimation. Alternatively, apply the augmentation to a dataset with independent features (no factor structure): if it still improves prediction, the mechanism is regularization rather than factor extraction; if it does not, factor structure is essential.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is empirical: for a fixed learning algorithm, replacing the raw feature matrix X with (F, U)—where F are K latent factors estimated from a transformed version of X (interactions, Gaussian or polynomial kernels, or an FNN hidden layer) and U is the residual of X after regressing on F—improves prediction. The paper demonstrates this on the Chinese news stock-return problem, where the FNN-based factor gives the best gains under Lasso, Ridge, and random forests, and the Gaussian-kernel factor ranks among the best under all five algorithms. The improvements transfer to downstream tasks: sentiment scores from augmented features produce sharper event-

Load-bearing premise

The load-bearing premise is that the estimated factors actually capture the latent factor space; in the main Chinese-news experiment the pretraining sample (n′=1,000) is smaller than the theory says is needed (about 2,400 for K≈15, p≈45,000), so the added factors may be too noisy.

Editorial extensions

If this is right

  • Augmentation is algorithm-agnostic: any regression or classification learner can be run on (F, U) with the same hyperparameters, so gains are additive to model improvements.
  • Nonlinear factors carry information beyond the linear factor F0; adding both F0 and a transformed factor usually improves on either alone.
  • The method needs only a handful of extra dimensions (K ≈ 12–15 here), so the variance cost is negligible relative to the high-dimensional X.
  • Diversified projection makes the framework feasible at web scale, since factor estimation uses a tiny pretraining subsample rather than PCA on the full n × p matrix.
  • Expected gain is contextual: the largest improvements appear when the baseline model is weak, while datasets with tiny samples or near-saturated baselines show little or no benefit.

Reading between the lines

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

  • Beyond the paper: because the FNN-based factor is fit with the response variable, part of its advantage likely reflects supervision leakage; an honest comparison would use a response-blind or cross-fitted feature extractor. The paper notes the supervision but does not quantify how much of the gain comes from it.
  • Beyond the paper: the success of kernel and interaction factors suggests the gain comes from exposing multiplicative structure that linear PCA on X misses; one testable extension is applying the same augmentation to genomic interaction or recommender data where pairwise effects are known to matter.
  • Beyond the paper: the method's logic may extend to intermediate network layers and tensor-valued designs; a direct experiment would take a pre-trained transformer's last hidden layer, extract factors, and see whether downstream fine-tuning improves.
  • Beyond the paper: a simple diagnostic for the mechanism is to permute the response when training the FNN factor; if out-of-sample gains persist under permutation, the benefit is regularization-like rather than genuinely predictive factor content.
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 a feature-augmentation framework for high-dimensional supervised learning. For a design matrix X, the method extracts latent factors from X and from nonlinear transformations of X (pairwise interactions, kernel matrices, and the last hidden layer of an FNN), then residualizes X on the extracted factors and feeds both the factors and the idiosyncratic residual matrix into a downstream learner. The authors apply the approach to stock-return prediction using Chinese financial news data and report out-of-sample R2 improvements for Lasso, Ridge, Random Forests, Gradient Boosted Trees, and Neural Networks. They also report event-study and portfolio analyses and present additional classification/regression experiments on MNIST, Fashion-MNIST, CIFAR-10, Reuters, bond risk premia, COVID-19, taxi demand, and Zillow data. The central claim is that augmenting the feature space with such factors consistently improves prediction performance for the same algorithm.

Significance. If the claim is correct, the paper offers a simple, modular, and potentially widely applicable preprocessing step that sits between data collection and algorithm design. The residualization algebra in Eqs. (2)-(4) is sound, and the empirical scope is unusually broad, spanning text, image, biological, financial, and epidemiological datasets. The supplementary experiments are a genuine strength and make the paper a useful reference point for practitioners. However, the central empirical claim is weakened by an evaluation protocol that selects the best of several augmentation variants after observing test performance, by the absence of uncertainty quantification, and by the inclusion of a supervised FNN factor in the comparisons. These issues are load-bearing because the paper's contribution is empirical rather than theoretical.

major comments (3)
  1. [§4.3, §5.3, Figures 2, 5, 9, 10] The headline claim of 'consistent improvement' is based on reporting the best of three augmentation schemes: (F,U), (F0,F,U), and (F0,F,eU). Section 5.3 further concludes only that 'for each dataset and algorithm, there exist augmentation methods that improve the estimation.' Because the augmentation rule is selected after inspecting test-set performance, the reported R2 ratios are maxima over a small menu and do not describe the performance of a fixed, pre-specified procedure. Although 20 repetitions are averaged, no confidence intervals, standard errors, or paired tests are reported. I request that the authors report all augmentation variants separately with dispersion measures and, ideally, a pre-specified rule for selecting the augmentation (e.g., by validation performance) so that the reported gains are not post-selection artifacts.
  2. [§3.2.2, §4.1] The diversified projection factor estimator is justified by the rate n' ≈ K^2 log p cited from Fan and Gu (2024). In the main Chinese-news application, the interaction matrix has p ≈ 44,850 (300 screened features, pairwise products), and the eigen-ratio method selects K ≈ 15, so K^2 log p ≈ 2,400. Yet Section 4.1 sets n' = 1,000. This violates the asymptotic sample-size requirement for accurate factor estimation. The authors should either justify why n' = 1,000 is sufficient in this application, increase n' to meet the rate, or provide a sensitivity analysis showing that the estimated factor space is stable across n' values. Without this, the main application's factor estimates may be too noisy to support the claimed gains.
  3. [§4.3, §5.3, §6] The paper explicitly admits in Section 4.3 that the FNN factor 'actually uses some information about the response variable (supervised), while all other methods do not.' The abstract and Section 6 claim that factors are 'extracted from design matrices and their transformations,' but a hidden layer trained on the response is not a design-matrix-only transformation. The strong results of Ffnn therefore do not support the unsupervised augmentation claim. I recommend reporting the unsupervised factors separately from the supervised FNN factor, or replacing the supervised FNN with an unsupervised autoencoder-based factor, so that the general claim is not carried by a supervised component.
minor comments (6)
  1. [§3.5] Typo: 'statistical machining learn methods' should be 'statistical machine learning methods.'
  2. [§4.3] The text says 'We consider five types of augmentation factors— F_inter, F_poly, F_rbf, and F_fnn' but lists only four named factors; the fifth appears to be F_0. Please clarify.
  3. [§5.2] The CIFAR-10 preprocessing cites 'Recommendation 602 from ITU-R', while the DogCat preprocessing cites 'Recommendation 601'; the correct ITU-R recommendation for grayscale conversion is BT.601. Please correct and ensure consistency.
  4. [Figure 4 caption] Typo: 'red, blud, green, purple, orange, and brown' should read 'red, blue, green, purple, orange, and brown.'
  5. [Table 1] Under 'Classification / L+S', the APR for (F0,U) is printed as '25,4' with a comma; should be '25.4'.
  6. [§5.3] The statement 'the proposed feature augmentation approaches usually tend to be powerful when the initial estimation does not have high accuracy' is a useful caveat, but it is stated after the fact. A pre-specified condition or a quantitative threshold for 'not high accuracy' would make this guidance more actionable.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: augmentation factors are estimated from design matrices and evaluated out-of-sample; self-citations are tools, not conclusions.

full rationale

The paper's claim is empirical rather than a derivation: it compares out-of-sample R2 or classification error with and without factors extracted from the original design matrix and its transformations. Factor estimation in Section 3.2 (PCA and diversified projection) uses only the design matrix; the test labels are not used to construct F0, F_inter, Frbf, or Fpoly. The subspace identity in Section 3.3, stating that (F_inter, U_inter) spans the same space as (F_inter, X), is a reparameterization, not a circular reduction: the augmented model still adds the factor directions, and the prediction is not algebraically forced. The only label-dependent factor is the FNN factor; the paper explicitly notes in Section 4.3 that 'the FNN factor actually uses some information about the response variable (supervised), while all other methods do not (unsupervised).' This is a fairness/leakage concern for the comparison, not a fitted parameter renamed as a prediction: the factor is an intermediate representation, and the reported R2 is out-of-sample. Self-citations (Fan–Gu 2024, Fan–Ke–Wang 2020, Fan–Liao 2022, Zhou–Fan–Xue 2024) supply estimation tools and the dataset; none is invoked as a uniqueness theorem or as the reason the augmentation works. A correctness risk exists but is not circular: diversified projection's consistency requires n' approximately K^2 log p (Section 3.2.2), while the main application sets n'=1000 with K≈15 and interaction dimension p≈45,000, failing the stated condition. Similarly, reporting 'the best outcomes among the three augmentations' (Sections 4.3 and 5.3) is selection-on-test, not circularity. No equation or fitted parameter reduces by construction to the input, so the derivation chain is self-contained.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central empirical claim rests on the factor-model assumption (Eq. 1), the consistency of the factor estimators (PCA and diversified projections), the lossless residualization in (2)-(4), and several heuristic choices (eigen-ratio bounds, kernel hyperparameters, n0, n', screening size) that are not all reported. The most fragile of these is the diversified projection pre-training size, which may violate the theoretical requirement in the main application.

free parameters (6)
  • Number of factors K = 12 for F0, 15 for Finter in the first window, chosen per window by eigen-ratio
    Determines the dimension of augmentation; chosen by a heuristic (eigen-ratio) rather than a formal test.
  • Pre-training sample size n' for diversified projection = 1000
    Set fixed for all windows; likely below the theoretical requirement n' ≈ K^2 log p for the interaction matrix.
  • Kernel hyperparameters (RBF bandwidth, polynomial degree) = not reported
    The paper does not state the bandwidth or degree used for the Gaussian and polynomial kernels, which materially affect the factors.
  • Number of randomly selected kernel columns n0 = not reported
    In Section 3.1, n0 columns are randomly selected to form the reduced kernel matrix; n0 is never specified.
  • Number of screened features (Chinese news) = 300
    The two-step reduction keeps the 3,000 most frequent words then screens to 300; this is a fixed choice.
  • Supervised FNN factor architecture = two hidden layers, widths 16 and 4 (for event-study classifier); architecture for X_fnn not specified
    The FNN used to extract X_fnn is not described; the event-study classifier uses widths 16 and 4.
assumptions (5)
  • domain assumption The data (or transformed data) follows an approximate factor model z_i = a + B f_i + u_i with f_i uncorrelated with u_i (Eq. 1).
    Invoked at the start of Section 3.2 for all matrices used to extract factors; if the factor structure fails, the extracted factors are not meaningful.
  • standard math PCA consistently estimates the factor space (Bai 2003) and diversified projections with n' ≈ K^2 log p are consistent (Fan and Gu 2024).
    Used in Section 3.2 to justify factor estimation; these are established results, but the sample size condition may not hold in the application.
  • standard math The residualization X = F B^T + U defines a lossless re-parameterization, so (F,U) spans the same subspace as (F,X).
    This is exact linear algebra used in Section 3.3 and is not in question.
  • ad hoc to paper The eigen-ratio heuristic with kmin = max(floor((p∧n)/10), 2) and kmax = floor((p∧n)/3) selects the true number of factors.
    The bounds are described as heuristic in Section 3.2.1; mis-specifying K changes the augmented features.
  • ad hoc to paper Randomly subsampling n0 columns of the kernel matrix preserves the geometry needed to extract predictive factors.
    Section 3.1 assumes the reduced kernel matrix retains the essential geometry, but no justification or n0 value is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Feature Augmentations for High-Dimensional Learning." pith.science (2026). https://pith.science/paper/WAIQ3KSL

@misc{pith2026250900232,
  author       = {Pith},
  title        = {Pith review of: Feature Augmentations for High-Dimensional Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WAIQ3KSL}},
  note         = {Machine review of arXiv:2509.00232}
}
read the original abstract

High-dimensional measurements are often correlated which motivates their approximation by factor models. This holds also true when features are engineered via low-dimensional interactions or kernel tricks. This often results in over parametrization and requires a fast dimensionality reduction. We propose a simple technique to enhance the performance of supervised learning algorithms by augmenting features with factors extracted from design matrices and their transformations. This is implemented by using the factors and idiosyncratic residuals which significantly weaken the correlations between input variables and hence increase the interpretability of learning algorithms and numerical stability. Extensive experiments on various algorithms and real-world data in diverse fields are carried out, among which we put special emphasis on the stock return prediction problem with Chinese financial news data due to the increasing interest in NLP problems in financial studies. We verify the capability of the proposed feature augmentation approach to boost overall prediction performance with the same algorithm. The approach bridges a gap in research that has been overlooked in previous studies, which focus either on collecting additional data or constructing more powerful algorithms, whereas our method lies in between these two directions using a simple PCA augmentation.

Figures

Figures reproduced from arXiv: 2509.00232 by the authors.

Figure 1
Figure 1. Scree plot for covariance matrices of the original [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Ratio of the out-of-sample R2 of each model to that without feature augmentation (R2 (X), benchmark) for Chinese news text dataset by diversified projection. The bars above the horizontal line at 1 indicate that the corresponding factor augmentation methods perform better than the benchmark. comparably to or significantly better than the linear factor Fb0. The relative gains vary across different machine learning mo… view at source ↗
Figure 3
Figure 3. Event Study on Beta-Adjusted Returns. The [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Cumulative Log2 Returns of Value-Weighted (VW) Portfolios. The top two panels display results based on regression-based score estimators, while the bottom two panels correspond to scores derived from binary classification. The red, blud, green, purple, orange, and brow…
Figure 5
Figure 5. Figure 5: Left column: Ratio of the classification error (ERR) of each model to that without [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Comparison between PCA performances of (Fb0,Fb, Ue ) and (Fb, Ub ). Left: Ratio of the classification error (ERR) of (Fb0,Fb, Ue ) to the ERR of its corresponding (Fb, Ub ), averaging over all the classification problems. For each factor group, from left to right are r…
Figure 7
Figure 7. Figure 7: Left: Difference between the relative errors of factor augmentations and likeli [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: One day ahead rolling window forecast for the daily COVID-19 new cases in Switzerland [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Ratio of the classification error (ERR) of each model to that without feature augmen [PITH_FULL_IMAGE:figures/full_fig_p026_9.png]
Figure 10
Figure 10. Figure 10: Ratio of the out-of-sample R2 of each model to that without feature augmentation (R2 (X), benchmark). The histograms are results for applying PCA on the whole data to estimate factors (cf. Section 3.2.1) while the dashed lines are the corresponding results for estimat…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 42 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if new.block note output fin.entry FUNCTION b...

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  4. [4]

    barticle [author] Ahn , Seung C S. C. Horenstein , Alex R A. R. ( 2013 ). Eigenvalue ratio test for the number of factors . Econometrica 81 1203--1227 . barticle

    work page 2013

  5. [5]

    ( 2003 )

    barticle [author] Bai , Jushan J. ( 2003 ). Inferential theory for factor models of large dimensions . Econometrica 71 135--171 . barticle

    work page 2003

  6. [6]

    Ng , Serena S

    barticle [author] Bai , Jushan J. Ng , Serena S. ( 2002 ). Determining the number of factors in approximate factor models . Econometrica 70 191--221 . barticle

    work page 2002

  7. [7]

    barticle [author] Balli , Hatice Ozer H. O. S rensen , Bent E B. E. ( 2013 ). Interaction effects in econometrics . Empirical Economics 45 583--603 . barticle

    work page 2013

  8. [8]

    ( 2001 )

    barticle [author] Breiman , Leo L. ( 2001 ). Random forests . Machine learning 45 5--32 . barticle

    work page 2001

Show all 47 references
  1. [9]

    , Iovanella , Antonio A

    barticle [author] Cerqueti , Roy R. , Iovanella , Antonio A. , Mattera , Raffaele R. Storani , Saverio S. ( 2024 ). Improving the explainability of autoencoder factors for commodities through forecast-based Shapley values . Scientific Reports 14 19622 . barticle

    2024

  2. [10]

    , Fan , Jianqing J

    barticle [author] Chen , Elynn E. , Fan , Jianqing J. Zhu , Xiaonan X. ( 2024 ). Factor augmented matrix regression . arXiv preprint arXiv:2405.17744 . barticle

    2024 arXiv

  3. [11]

    , Yang , Dan D

    barticle [author] Chen , Rong R. , Yang , Dan D. Zhang , Cun-Hui C.-H. ( 2022 ). Factor models for high-dimensional tensor time series . Journal of the American Statistical Association 117 94--116 . barticle

    2022

  4. [12]

    barticle [author] Cochrane , John H J. H. Piazzesi , Monika M. ( 2005 ). Bond risk premia . American economic review 95 138--160 . barticle

    2005

  5. [13]

    barticle [author] Fama , Eugene F E. F. French , Kenneth R K. R. ( 1992 ). The cross-section of expected stock returns . the Journal of Finance 47 427--465 . barticle

    1992

  6. [14]

    Gu , Yihong Y

    barticle [author] Fan , Jianqing J. Gu , Yihong Y. ( 2024 ). Factor Augmented Sparse Throughput Deep ReLU Neural Networks for High Dimensional Regression . Journal of Americal Statistical Association . barticle

    2024

  7. [15]

    , Guo , Jianhua J

    barticle [author] Fan , Jianqing J. , Guo , Jianhua J. Zheng , Shurong S. ( 2022 ). Estimating number of factors by adjusted eigenvalues thresholding . Journal of the American Statistical Association 117 852--861 . barticle

    2022

  8. [16]

    , Ke , Yuan Y

    barticle [author] Fan , Jianqing J. , Ke , Yuan Y. Wang , Kaizheng K. ( 2020 ). Factor-adjusted regularized model selection . Journal of econometrics 216 71--85 . barticle

    2020

  9. [17]

    Liao , Yuan Y

    barticle [author] Fan , Jianqing J. Liao , Yuan Y. ( 2022 ). Learning latent factors from diversified projections and its applications to over-estimated and weak factors . Journal of the American Statistical Association 117 909--924 . barticle

    2022

  10. [18]

    Lv , Jinchi J

    barticle [author] Fan , Jianqing J. Lv , Jinchi J. ( 2008 ). Sure independence screening for ultrahigh dimensional feature space . Journal of the Royal Statistical Society: Series B (Statistical Methodology) 70 849--911 . barticle

    2008

  11. [19]

    , Feng , Yang Y

    barticle [author] Fan , Jianqing J. , Feng , Yang Y. , Jiang , Jiancheng J. Tong , Xin X. ( 2016 ). Feature augmentation via nonparametrics and selection (FANS) in high-dimensional classification . Journal of the American Statistical Association 111 275--287 . barticle

    2016

  12. [20]

    , Spatt , Chester S C

    barticle [author] Goldstein , Itay I. , Spatt , Chester S C. S. Ye , Mao M. ( 2021 ). Big data in finance . The Review of Financial Studies 34 3213--3225 . barticle

    2021

  13. [21]

    , Oyallon , Edouard E

    barticle [author] Grinsztajn , L \'e o L. , Oyallon , Edouard E. Varoquaux , Ga \"e l G. ( 2022 ). Why do tree-based models still outperform deep learning on typical tabular data? Advances in neural information processing systems 35 507--520 . barticle

    2022

  14. [22]

    , Gardiner , Katheleen J K

    barticle [author] Higuera , Clara C. , Gardiner , Katheleen J K. J. Cios , Krzysztof J K. J. ( 2015 ). Self-organizing feature maps identify proteins critical to learning in a mouse model of down syndrome . PloS one 10 e0129126 . barticle

    2015

  15. [23]

    barticle [author] Hinton , Geoffrey E G. E. , Srivastava , Nitish N. , Krizhevsky , Alex A. , Sutskever , Ilya I. Salakhutdinov , Ruslan R R. R. ( 2012 ). Improving neural networks by preventing co-adaptation of feature detectors . arXiv preprint arXiv:1207.0580 . barticle

    2012 arXiv

  16. [24]

    btechreport [author] Ke , Zheng Tracy Z. T. , Kelly , Bryan T B. T. Xiu , Dacheng D. ( 2019 ). Predicting returns with text data Technical Report , National Bureau of Economic Research . btechreport

    2019

  17. [25]

    , Hinton , Geoffrey G

    barticle [author] Krizhevsky , Alex A. , Hinton , Geoffrey G. et al. ( 2009 ). Learning multiple layers of features from tiny images . barticle

    2009

  18. [26]

    Yao , Qiwei Q

    barticle [author] Lam , Clifford C. Yao , Qiwei Q. ( 2012 ). Factor modeling for high-dimensional time series: inference for the number of factors . The Annals of Statistics 694--726 . barticle

    2012

  19. [27]

    ( 1998 )

    barticle [author] LeCun , Yann Y. ( 1998 ). The MNIST database of handwritten digits . http://yann. lecun. com/exdb/mnist/ . barticle

    1998

  20. [28]

    McDonald , Bill B

    barticle [author] Loughran , Tim T. McDonald , Bill B. ( 2016 ). Textual analysis in accounting and finance: A survey . Journal of Accounting Research 54 1187--1230 . barticle

    2016

  21. [29]

    , Wijaya , Adi A

    barticle [author] Machmud , Rizanda R. , Wijaya , Adi A. et al. ( 2016 ). Behavior determinant based cervical cancer early detection with machine learning algorithm . Advanced Science Letters 22 3120--3123 . barticle

    2016

  22. [30]

    Pastur , Leonid A L

    barticle [author] Marchenko , VA V. Pastur , Leonid A L. A. ( 1967 ). Distribution of eigenvalues for some sets of random matrices . Mat. Sb.(NS) 72 4 . barticle

    1967

  23. [31]

    barticle [author] McCracken , Michael W M. W. Ng , Serena S. ( 2016 ). FRED-MD: A monthly database for macroeconomic research . Journal of Business & Economic Statistics 34 574--589 . barticle

    2016

  24. [32]

    , Mathieu , Edouard E

    barticle [author] Ritchie , Hannah H. , Mathieu , Edouard E. , Rodes-Guirao , Lucas L. , Appel , Cameron C. , Giattino , Charlie C. , Ortiz-Ospina , Esteban E. , Hasell , Joe J. , Macdonald , Bobbie B. , Beltekian , Diana D. Roser , Max M. ( 2020 ). Coronavirus Pandemic (COVID...

    2020

  25. [33]

    , Markou , Ioulia I

    barticle [author] Rodrigues , Filipe F. , Markou , Ioulia I. Pereira , Francisco C F. C. ( 2019 ). Combining time-series and textual data for taxi demand prediction in event areas: A deep learning approach . Information Fusion 49 120--129 . barticle

    2019

  26. [34]

    Armon , Amitai A

    barticle [author] Shwartz-Ziv , Ravid R. Armon , Amitai A. ( 2022 ). Tabular data: Deep learning is not all you need . Information Fusion 81 84--90 . barticle

    2022

  27. [35]

    barticle [author] Stock , James H J. H. Watson , Mark W M. W. ( 2002 ). Macroeconomic forecasting using diffusion indexes . Journal of Business & Economic Statistics 20 147--162 . barticle

    2002

  28. [36]

    ( 2017 )

    bmisc [author] Sun , Junyi J. ( 2017 ). Jieba: Chinese text segmentation . https://github.com/fxsjy/jieba . Accessed: April 2025 . bmisc

    2017

  29. [37]

    , Lin , Sue-Jane S.-J

    barticle [author] Tsai , Shu-Chun S.-C. , Lin , Sue-Jane S.-J. , Chen , Po-Wen P.-W. , Luo , Wen-Yi W.-Y. , Yeh , Te-Huei T.-H. , Wang , Hsei-Wei H.-W. , Chen , Chi-Ju C.-J. Tsai , Ching-Hwa C.-H. ( 2009 ). EBV Zta protein induces the expression of interleukin-13, promoting th...

    2009

  30. [38]

    bbook [author] Wainwright , Martin J M. J. ( 2019 ). High-dimensional statistics: A non-asymptotic viewpoint 48 . Cambridge university press . bbook

    2019

  31. [39]

    , Liu , Xialu X

    barticle [author] Wang , Dong D. , Liu , Xialu X. Chen , Rong R. ( 2019 ). Factor models for matrix-valued high-dimensional time series . Journal of econometrics 208 231--248 . barticle

    2019

  32. [40]

    , Cheng , Yu Y

    barticle [author] Wang , Liyang L. , Cheng , Yu Y. , Xiang , Ao A. , Zhang , Jingyu J. Yang , Haowei H. ( 2024 ). Application of natural language processing in financial risk detection . arXiv preprint arXiv:2406.09765 . barticle

    2024 arXiv

  33. [41]

    , Zhang , Xiao-Fei X.-F

    barticle [author] Wu , Meng-Yun M.-Y. , Zhang , Xiao-Fei X.-F. , Dai , Dao-Qing D.-Q. , Ou-Yang , Le L. , Zhu , Yuan Y. Yan , Hong H. ( 2016 ). Regularized logistic regression with network-based pairwise interaction for biomarker identification in breast cancer . BMC bioinform...

    2016

  34. [42]

    , Rasul , Kashif K

    barticle [author] Xiao , Han H. , Rasul , Kashif K. Vollgraf , Roland R. ( 2017 ). Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms . arXiv preprint arXiv:1708.07747 . barticle

    2017 arXiv

  35. [43]

    Shen , Zhouyu Z

    barticle [author] Xiu , Dacheng D. Shen , Zhouyu Z. ( 2024 ). Deep Autoencoders for Nonlinear Factor Models: Theory and Applications . Available at SSRN . barticle

    2024

  36. [44]

    , Zhou , Wen W

    barticle [author] Zhang , Lyuou L. , Zhou , Wen W. Wang , Haonan H. ( 2022 ). Non-asymptotic properties of spectral decomposition of large Gram-type matrices and applications . Bernoulli 28 1224--1249 . barticle

    2022

  37. [45]

    , Zhang , Hao H

    binproceedings [author] Zhang , Die D. , Zhang , Hao H. , Zhou , Huilin H. , Bao , Xiaoyi X. , Huo , Da D. , Chen , Ruizhao R. , Cheng , Xu X. , Wu , Mengyue M. Zhang , Quanshi Q. ( 2021 ). Building interpretable interaction trees for deep nlp models . In Proceedings of the AA...

    2021

  38. [46]

    , Fan , Jianqing J

    barticle [author] Zhou , Yang Y. , Fan , Jianqing J. Xue , Lirong L. ( 2024 ). How Much Can Machines Learn Finance from Chinese Text Data? Management Science . barticle

    2024

  39. [47]

    , Xue , Lirong L

    barticle [author] Zhou , Yang Y. , Xue , Lirong L. , Shi , Zhengyu Z. , Wu , Libo L. Fan , Jianqing J. ( 2023 ). Measuring Housing Activeness from Multi-Source Big Data and Machine Learning . Journal of American Statistical Association 117 1045-1059 . barticle

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.