Pith. sign in

REVIEW 4 major objections 4 minor 22 references

PH-VAE: A Polynomial Hierarchical Variational Autoencoder Towards Disentangled Representation Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A polynomial hierarchical variational autoencoder significantly improves reconstruction and generation accuracy while adding a concrete form of disentanglement.

desk verdict The paper's central ELBO identity is mathematically false, so the theoretical claims collapse; the experiments don't compensate. read the letter →

arxiv 2502.02856 v2 pith:DRWSXUNC submitted 2025-02-05 cs.LG

Xi Chen , Shaofan Li This is my paper · ORCID
classification cs.LG
keywords variationalautoencoderpolynomialdivergencedisentangledrepresentationlearninghierarchicallatentvariablemodelgenerativeartificialintelligenceimagereconstructionKullback-Leiblerfeatures
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a variational autoencoder can reconstruct and generate complex data distributions more accurately and more reproducibly if the input is first expanded into polynomial powers $x, x^2, x^3, \ldots$, with each power processed by its own encoder and all encoders sharing one decoder. It replaces the single Kullback-Leibler divergence in the VAE loss with a Polynomial Divergence, the average of per-layer KL divergences, and argues this generalization preserves a valid evidence lower bound. On uniform and log-normal toy data, on two known VAE failure cases, and on MNIST and a 10-image CelebA subset, the authors report that adding polynomial features systematically lowers the loss and sharpens reconstructions as $S$ grows, while the noise amplitude $A$ enlarges the latent exploration. The paper further claims that the hierarchical posterior mixture provides a concrete, measurable form of disentangled representation learning.

What carries the argument

The central object is the polynomial hierarchical data format: each sample $x$ is expanded into powers $x^1, \ldots, x^S$, and each power is fed to a separate encoder that outputs its own mean $\mu_s$ and variance $\sigma_s^2$. These are averaged to give the latent $z = \mu + A \cdot \epsilon \cdot \sigma$, where $A$ is a tunable noise amplifier. The loss uses the Polynomial Hierarchical Divergence, $\mathrm{PH}(q(z|x)\|p(z)) = \frac{1}{S} \sum_{s=1}^S \mathrm{KL}[q_s(z|x^s)\|p(z)]$, in place of a single KL term; with Gaussian $q_s$ it has the closed form $\frac{1}{2S} \sum_{s=1}^S (1 + \log \sigma_s^2 - \mu_s^2 - \sigma_s^2)$. The assertion that this is a valid generalization of KL divergence, and the corresponding ELBO derivation, is what carries the claimed improvements.

What would settle it

On the Gaussian-mixture example with known true density $p(x)$, compute the exact marginal log-likelihood $\log p(x)$ and the PH-VAE training objective $L^{\mathrm{PH}}_{\mathrm{ELBO}}$ after convergence; a valid ELBO must satisfy $L^{\mathrm{PH}}_{\mathrm{ELBO}} \le \log p(x)$. Any batch where the computed objective exceeds $\log p(x)$ would falsify the derivation in Eqs. (11)-(17). A second check is to train PH-VAE with $S=1$ and with $S=3$ on identical random seeds and verify that the $S=3$ model's reconstruction error is lower across many initializations, since the paper's central improvement claim is monotonic in $S$.

Watch

Extended reading notes

Core claim

The central claim is that the standard VAE's reconstruction and generation failures come from compressing all information through a single latent bottleneck with one KL penalty, and that this can be remedied by a polynomial hierarchical input format. PH-VAE models the joint distribution as $p(x,z) \approx p(x|z) \cdot \frac{1}{S} \sum_{s=1}^S p(z|x^s)$, approximates the posterior by the equal-weight mixture $q(z|x) = \frac{1}{S} \sum_{s=1}^S q_s(z|x^s)$, and trains with the objective $L^{\mathrm{PH}}_{\mathrm{ELBO}} = \mathbb{E}_{q(z|x)}[\log p(x|z)] - \frac{1}{S} \sum_{s=1}^S \mathrm{KL}[q_s(z|x^s) \| p(z)]$. The authors assert that this objective is an ELBO and that the averaged KL penalty, which they call the Polynomial Hierarchical Divergence, is a generalized divergence. The claimed payoff is that higher-order polynomial features inject multi-scale information into the latent space without increasing dataset size, reducing posterior collapse and yielding what the authors describe as systematic and drastic improvements in accuracy and reproducibility.

Load-bearing premise

The load-bearing premise is that $p(x,z) \approx p(x|z)$ times the average of $p(z|x^s)$, with the equal-weight mixture $q(z|x) = \frac{1}{S} \sum_s q_s(z|x^s)$ serving as a valid variational posterior, a step the paper states follows from the IWAE metric without giving the derivation; if this identity does not hold, the ELBO and the mutual-information improvement claimed for PH-VAE collapse.

Editorial extensions

If this is right

  • For a fixed dataset size, reconstruction accuracy should improve monotonically as $S$ increases, because each added polynomial feature adds a distinct scale of latent information.
  • Increasing the noise amplifier $A$ should not blur reconstructions away from the data; PH-VAE is claimed to remain accurate for $A=3$ and $A=5$ where VAE degrades.
  • The averaged KL penalty should prevent latent collapse, keeping the mutual information between data and latent higher than in a standard VAE.
  • The same architecture should transfer to other data modalities, such as time series or molecular data, because the polynomial expansion is modality-agnostic.
  • Reproducibility of reconstructed distributions should improve in the sense that repeated runs produce similar density estimates, as suggested by the 100-run averaging protocol in Example 1.

Reading between the lines

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

  • If the equal-weight mixture posterior is the right reading, then PH-VAE is effectively an ensemble of $S$ variational posteriors tied to polynomial views of the same input; this connects to multi-view or data-augmentation VAEs, a connection the paper does not make.
  • The polynomial divergence's status as a generalized divergence hinges on each $q_s$ being a valid normalized density; if any encoder output drifts from normalization during training, the nonnegativity and ELBO claims would need explicit checking.
  • A direct testable extension is to apply PH-VAE to a benchmark with known ground-truth generative factors, such as dSprites or Shapes3D, and measure standard disentanglement metrics to see whether the claimed disentanglement is competitive with existing methods.
  • The paper's reproducibility claim could be quantified by computing distributional distances, such as Wasserstein or Kolmogorov-Smirnov, between repeated PH-VAE reconstructions and ground truth, rather than relying on visual inspection.
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 / 4 minor

Summary. The manuscript proposes PH-VAE, a variational autoencoder variant that feeds polynomial powers x^s of the input into separate encoders, averages the resulting latent means and variances, and uses a loss in which the KL regularization is replaced by an average of per-encoder KL divergences to a prior, called the Polynomial Hierarchical Divergence. The authors claim that this construction improves reconstruction accuracy and reproducibility, mitigates posterior collapse, and provides unsupervised disentangled representation learning. The theoretical claims are developed in Section 2.3 and 2.4, and the experimental section reports qualitative comparisons on synthetic distributions, MNIST, and CelebA using only visual inspection of reconstructed images.

Significance. If the central derivation were sound, the proposed architecture would be a simple, easily implementable modification of the standard VAE with a concrete objective, and the code snippet in Section 2.3 is a useful starting point for reproducing the method. However, the load-bearing theoretical steps are incorrect: Eq. (18) does not follow from the preceding definitions, and Eq. (22) misidentifies an average of per-encoder KL terms as mutual information. The experimental section contains no quantitative metrics, no held-out evaluation, and no error bars, so the claimed 'systematic and drastic improvements' are not supported by the reported evidence. The paper does not provide machine-checked proofs or a complete reproducible artifact beyond an illustrative code fragment.

major comments (4)
  1. [2.3, Eq. (18)] The derivation of Eq. (18) is invalid. Starting from the standard identity log p(x) = E_{q(z|x)}[log p(x|z)] - KL(q(z|x)||p(z)) + KL(q(z|x)||p(z|x)), substituting the PH-ELBO in Eq. (13) gives L_PH = log p(x) - KL(q(z|x)||p(z|x)) - [(1/S)Σ_s KL(q_s(z|x^s)||p(z)) - KL(q(z|x)||p(z))]. The bracketed term is generally nonzero because the paper explicitly constructs distinct encoders and notes q(z|x) ≠ q_1(z|x). Thus Eq. (18) is false in general, and the proposed objective is not a valid lower bound on log p(x) under the stated model. This invalidates the theoretical justification for the polynomial divergence loss.
  2. [2.4, Eqs. (22)-(23)] The quantity I(x,z) defined in Eq. (23) as (1/S)Σ_{s=2}^S KL[q_s(z|x^s)||p(z)] is not mutual information. The mutual information between x and z is E_{p(x)}[KL[p(z|x)||p(z)]], not an average of per-encoder KL divergences to the prior. Moreover, Eq. (22) merely relabels the KL terms already present in Eq. (13): the first KL term is kept and the remaining KL terms are renamed I(x,z). Consequently, the claim that PH-VAE 'improves mutual information' is true by construction rather than by an independent decomposition, and it does not provide evidence of disentanglement.
  3. [2.3, Eq. (11)] The joint model p(x,z) ≈ p(x|z) · (1/S)Σ_s p(z|x^s) is stated without a derivation. The text says it follows from the IWAE metric, but no such derivation is given, and the expression is not obviously a properly normalized joint distribution. This premise is load-bearing because it motivates the averaged posterior q(z|x) = (1/S)Σ_s q_s(z|x^s) in Eq. (12) and hence the form of the loss. Without a valid derivation, the theoretical foundation of the method is incomplete.
  4. [3, Experiments] The experimental evaluation is exclusively qualitative. There are no quantitative metrics such as reconstruction error, negative log-likelihood, FID, or disentanglement scores; there is no held-out test set described for the image experiments; and no error bars or repeated-run statistics are reported. The MNIST and CelebA experiments use only 10 images each, and the figures show reconstructed training images. These results cannot support the abstract's claims of 'systematic and drastic improvements in both accuracy and reproducibility' or the conclusion that PH-VAE 'achieves superior reconstruction accuracy and generative ability.'
minor comments (4)
  1. [Abstract] The phrase 'polynomial hierarchical date format' should be 'data format'.
  2. [2.4] The text refers to 'ELBO objective in Eq. (14)', but Eq. (14) defines the KL divergence; the ELBO is defined in Eq. (13).
  3. [3.1] The text says 'the activation functions g(·) in Eq. (4) and f(·) in Eq. (8)' but Eq. (4) is the polynomial data transformation and Eq. (8) is the log-variance formula; the activation functions appear in Eqs. (6) and (10).
  4. [2.3, code snippet] In the loss_function code, kl_loss3 is used in the return statement but is not defined in the function body, which prevents the snippet from running as shown.

Circularity Check

2 steps flagged · score 7.0 of 10

The claimed MI/disentanglement improvement is an algebraic identity: Eq. (22) with I defined in Eq. (23) is simply Eq. (13), so PH-VAE's central theoretical benefit is a relabeling of its own KL terms rather than an independent prediction.

  1. self definitional [Section 2.4, Eqs. (22)-(23)]
    "According to (Mutual Information) MI decomposition, ELBO objective in Eq. (14) can be modified as, L^PH_ELBO = E_{q(z|x)}[log p(x|z)] − I(x, z) − (1/S) KL[q_1(z|x)∥p(z)], (22) where I(x, z) := (1/S) Σ_{s=2}^S KL[q_s(z|x^s)∥p(z)], (23) where I(x, z) is the mutual information (MI)."

    Substituting Eq. (23) into Eq. (22) yields exactly the loss defined in Eq. (13): E_q[log p(x|z)] − (1/S)Σ_{s=1}^S KL[q_s||p(z)] = E_q[log p(x|z)] − (1/S)KL[q_1||p(z)] − I(x,z). Hence I(x,z) is not a standard mutual information such as E[KL(q(z|x)||q(z))]; it is the s≥2 slice of the paper's own average-KL regularizer. The conclusion that PH-VAE improves MI is therefore true by construction: the loss explicitly penalizes those KL terms, and their reduction is reported as MI improvement. The disentanglement claim is a relabeling of the loss components, not a derived prediction.

  2. renaming known result [Section 2.3, Eqs. (14)-(15)]
    "We define the average of the Kullback-Leibler (KL) Divergences of all layers as the Polynomial Hierarchical Divergence, i.e. PH(q(z|x)∥p(z)) := 1/S Σ_{s=1}^S KL [q_s(z|x^s)∥p(z)] ."

    The Polynomial Hierarchical Divergence is defined as an arithmetic average of ordinary KL divergences between each component posterior and the prior. Averaging a known divergence does not produce a new divergence measure; it is a convex combination of the same KL quantities used in standard VAEs. The paper's presentation of PH as a novel divergence that generalizes KL is thus a renaming of the existing regularization, and the theoretical explanation of the reported gains reduces to the properties of the standard KL term rather than to a new mathematical object.

full rationale

The strongest circularity is in Section 2.4. Eq. (22) presents a mutual information decomposition, but I(x,z) defined in Eq. (23) is exactly the s=2..S portion of the average-KL regularizer already present in Eq. (13). Substituting (23) into (22) recovers (13) algebraically, so the statement that PH-VAE improves MI is an identity, not an independent result. The disentangled parts of higher order scale information are therefore the paper's own KL penalties under another name. Separately, Section 2.3 asserts without proof that Eq. (11) follows from the IWAE metric, and Eqs. (17)-(18) do not constitute a valid ELBO derivation: with distinct q_s, L^PH in Eq. (13) is not equal to log p(x) - (1/S)Σ KL[q_s||p(z|x)] unless a mixture-KL bracket that generally does not vanish is zero. These are correctness defects rather than circularities, but they mean the theoretical foundation is not independently established. The image and probability reconstruction experiments are empirical comparisons rather than circular predictions, though they are confounded by increased model capacity and lack quantitative metrics, held-out evaluation, or error bars. There is no load-bearing self-citation; the main problem is definitional and derivational, not a citation loop.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claim rests on a small number of hand-chosen parameters (A, S, equal mixture weights), an unproven joint model, an unjustified mixture posterior, and a relabeling of KL terms as mutual information. The paper introduces no independent evidence for these assumptions.

free parameters (3)
  • A (noise amplifier) = 1, 3, and 5
    Hand-chosen scaling of the reparameterization noise in Eq. (9); it directly changes the sampled latent z and the reported results, but no selection criterion or ablation is given.
  • S (number of polynomial feature levels) = 3 for toy data, 5 for MNIST and CelebA
    The core architectural choice: adding x^2 through x^S streams is the claimed source of improvement; S is chosen ad hoc, and Figures 7 and 9 are effectively a sweep over S with no statistical validation.
  • Equal mixture weight 1/S = 1/S
    The posterior q(z|x) and the PH divergence weight each encoder equally; this is an ad hoc choice that determines the loss and the claimed MI decomposition.
assumptions (4)
  • ad hoc to paper Joint model p(x,z) ≈ p(x|z) * (1/S) Σ_s p(z|x^s)
    Stated in Eq. (11) as derivable from IWAE, but no derivation is given and the expression is not shown to be a normalized joint distribution.
  • ad hoc to paper Variational posterior q(z|x) = (1/S) Σ_s q_s(z|x^s)
    Eq. (12) defines the surrogate posterior as an equal-weight mixture of per-encoder Gaussians; this mixture choice is unjustified and makes the ELBO derivation in Eqs. (17)-(18) invalid.
  • domain assumption Gaussian closed-form KL with factorized posterior
    Eq. (20) assumes each q_s is a factorized Gaussian with diagonal covariance; the paper itself notes that the product of two Gaussian variables is not Gaussian, so the closed form is an approximation.
  • ad hoc to paper Equation (22): I(x,z) := (1/S) Σ_{s=2}^S KL[q_s(z|x_s)||p(z)] is mutual information
    This is asserted in Section 2.4. The defined quantity is a weighted sum of KL penalties, not the mutual information I_q(x;z) = KL[q(z|x)||q(z)], so the MI-improvement conclusion is built into a relabeling.
invented entities (1)
  • Polynomial Divergence PH(q||p)
    purpose: Replacement for KL divergence in the loss and claimed tool for measuring and controlling disentanglement.
    Eq. (15) defines it as the arithmetic mean of the standard per-layer KL divergences; it is not an independent divergence and has no falsifiable prediction outside the loss it is built from.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PH-VAE: A Polynomial Hierarchical Variational Autoencoder Towards Disentangled Representation Learning." pith.science (2026). https://pith.science/paper/DRWSXUNC

@misc{pith2026250202856,
  author       = {Pith},
  title        = {Pith review of: PH-VAE: A Polynomial Hierarchical Variational Autoencoder Towards Disentangled Representation Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DRWSXUNC}},
  note         = {Machine review of arXiv:2502.02856}
}
read the original abstract

The variational autoencoder (VAE) is a simple and efficient generative artificial intelligence method for modeling complex probability distributions of various types of data, such as images and texts. However, it suffers some main shortcomings, such as lack of interpretability in the latent variables, difficulties in tuning hyperparameters while training, producing blurry, unrealistic downstream outputs or loss of information due to how it calculates loss functions and recovers data distributions, overfitting, and origin gravity effect for small data sets, among other issues. These and other limitations have caused unsatisfactory generation effects for the data with complex distributions. In this work, we proposed and developed a polynomial hierarchical variational autoencoder (PH-VAE), in which we used a polynomial hierarchical date format to generate or to reconstruct the data distributions. In doing so, we also proposed a novel Polynomial Divergence in the loss function to replace or generalize the Kullback-Leibler (KL) divergence, which results in systematic and drastic improvements in both accuracy and reproducibility of the re-constructed distribution function as well as the quality of re-constructed data images while keeping the dataset size the same but capturing fine resolution of the data. Moreover, we showed that the proposed PH-VAE has some form of disentangled representation learning ability.

Figures

Figures reproduced from arXiv: 2502.02856 by the authors.

Figure 1
Figure 1. Architecture of the polynomial-enriched hierarchical variational autoencoder. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Probability densities of the predicted results for PH-VAE, VAE and the ground truth samples. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Values of the loss function in each epoch for the polynomial features [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Comparison of the ground truth data and the reconstructed results from VAE and PHVAE in the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Comparison of the ground truth data and the corresponding results obtained from VAE and [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Comparison of the ground truth data with the corresponding results obtained from VAE and [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Digital images for VAE, PH-VAE and the ground truth samples. (a) [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Values of the loss function in each epoch for the polynomial features [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Facial image comparison among VAE, PH-VAE and the ground truth samples. [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 15 canonical work pages

  1. [1]

    The autoencoding variational autoencoder

    Cemgil T, Ghaisas S, Dvijotham K, Gowal S, Kohli P. The autoencoding variational autoencoder. Advances in Neural Information Processing Systems. 2020;33:15077-15087

    work page 2020

  2. [2]

    Recent advances in variational autoencoders with representation learning for biomedical informatics: A survey

    Wei R, Mahmood A. Recent advances in variational autoencoders with representation learning for biomedical informatics: A survey. IEEE Access, 9:4939–4956. 2020

    work page 2020

  3. [3]

    Recent research and applications in variational autoencoders for industrial prognosis and health management: A survey

    Zemouri R, Lévesque M, Boucher É, Kirouac M, Lafleur F, Bernier S, Merkhouf A. Recent research and applications in variational autoencoders for industrial prognosis and health management: A survey. In: 2022 Prognostics and Health Management Conference (PHM-2022 London). IEEE; 2022 May. p. 193-203

    work page 2022

  4. [4]

    An overview of variational autoencoders for source separation, finance, and bio-signal applications

    Singh A, Ogunfunmi T. An overview of variational autoencoders for source separation, finance, and bio-signal applications. Entropy. 2021 Jan;24(1):55

    work page 2021

  5. [5]

    and Shi, S., Generative artificial intelligence and its applications in materials science: Current situation and future perspectives

    Liu, Y., Yang, Z., Yu, Z., Liu, Z., Liu, D., Lin, H., Li, M., Ma, S., Avdeev, M. and Shi, S., Generative artificial intelligence and its applications in materials science: Current situation and future perspectives. Journal of Materiomics, 9(4), pp.798-816. 2023

    work page 2023

  6. [6]

    and Lerchner, A., beta-VAE: Learning basic visual concepts with a constrained variational framework

    Higgins, I., Matthey, L., Pal, A., Burgess, C.P., Glorot, X., Botvinick, M.M., Mohamed, S. and Lerchner, A., beta-VAE: Learning basic visual concepts with a constrained variational framework. ICLR (Poster),

    work page

  7. [7]

    and Lerchner, A., Under- standing disentangling inβ-VAE

    Burgess, C.P., Higgins, I., Pal, A., Matthey, L., Watters, N., Desjardins, G. and Lerchner, A., Under- standing disentangling inβ-VAE. arXiv preprint arXiv:1804.03599. 2018

    arXiv 2018

  8. [8]

    and Kautz, J., NVAE: A deep hierarchical variational autoencoder

    Vahdat, A. and Kautz, J., NVAE: A deep hierarchical variational autoencoder. Advances in neural information processing systems, 33, pp.19667-19679. 2020

    work page 2020

Show all 22 references
  1. [9]

    and Balakrishnan, A

    Kumar, A., Sattigeri, P. and Balakrishnan, A. Variational inference of disentangled latent concepts from unlabeled observations. arXiv preprint arXiv:1711.00848, 2017

    2017 arXiv

  2. [10]

    and Ermon, S

    Zhao, S., Song, J. and Ermon, S. Infovae: Balancing learning and inference in variational autoencoders. In Proceedings of the AAAI conference on artificial intelligence 33(01), pp. 5885-5892. July, 2019

    2019

  3. [11]

    and Cho, S., Variational autoencoder based anomaly detection using reconstruction probability

    An, J. and Cho, S., Variational autoencoder based anomaly detection using reconstruction probability. Special lecture on IE, 2(1), pp.1-18. 2015

    2015

  4. [12]

    and Mnih, A., 2018, July

    Kim, H. and Mnih, A., 2018, July. Disentangling by factorising. In International conference on machine learning (pp. 2649-2658). PMLR

    2018

  5. [13]

    and Pavlovic, V., 2019

    Kim, M., Wang, Y., Sahu, P. and Pavlovic, V., 2019. Relevance factor vae: Learning and identifying disentangled factors. arXiv preprint arXiv:1902.01568

    2019 arXiv

  6. [14]

    and Roy, N., Task-conditioned variational autoencoders for learning movement primitives

    Noseworthy, M., Paul, R., Roy, S., Park, D. and Roy, N., Task-conditioned variational autoencoders for learning movement primitives. In Conference on robot learning (pp. 933-944). PMLR. 2020, May

    2020

  7. [15]

    and Alameda-Pineda, X., Dynamical variational autoencoders: A comprehensive review

    Girin, L., Leglaive, S., Bie, X., Diard, J., Hueber, T. and Alameda-Pineda, X., Dynamical variational autoencoders: A comprehensive review. arXiv preprint arXiv:2008.12595. 2020

    2008 arXiv

  8. [16]

    Failure modes of variational autoencoders and their effects on downstream tasks

    Yacoby Y, Pan W, Doshi-Velez F. Failure modes of variational autoencoders and their effects on downstream tasks. arXiv:2007.07124. 2020 Jul 14

    2007 arXiv

  9. [17]

    and Zhu, W., Disentangled representation learning

    Wang, X., Chen, H., Wu, Z. and Zhu, W., Disentangled representation learning. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2024

    2024

  10. [18]

    arXiv preprint arXiv:2211.07700

    Pastrana, R., Disentangling Variational Autoencoders. arXiv preprint arXiv:2211.07700. 2022

    2022 arXiv

  11. [19]

    and Dauwels, J.,α TC-VAE: On the relationship between Disentanglement and Diversity

    Meo, C., Mahon, L., Goyal, A. and Dauwels, J.,α TC-VAE: On the relationship between Disentanglement and Diversity. In The Twelfth International Conference on Learning Representations. 2024

    2024

  12. [20]

    Auto-Encoding Variational Bayes

    Kingma DP, Welling M. Auto-Encoding Variational Bayes. arXiv:1312.6114. 2014 Dec 20

    2014 arXiv

  13. [21]

    Tutorial on Variational Autoencoders

    Doersch C. Tutorial on Variational Autoencoders. arXiv:1606.05908. 2016 Jun 20

    2016 arXiv

  14. [22]

    Deep Learning Face Attributes in the Wild, In Proceedings of International Conference on Computer Vision (ICCV), December, 2015

    Liu, Z., Luo, P., Wang, X., and Tang, X. Deep Learning Face Attributes in the Wild, In Proceedings of International Conference on Computer Vision (ICCV), December, 2015. 20

    2015

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.