Pith. sign in

REVIEW 3 major objections 6 minor 40 references

P2L: Predicting Transfer Learning for Images and Semantic Relations

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Cheap score picks best transfer source in 13 of 21 tasks

desk verdict A held-out-tested estimator for transfer source selection, with a reference-model confound that limits but does not destroy the claim. read the letter →

arxiv 1908.07630 v2 pith:ZAMXNMWC submitted 2019-08-20 cs.LG cs.AIcs.CV

Bishwaranjan Bhattacharjee , John R. Kender , Matthew Hill , Parijat Dube , Siyu Huo , Michael R. Glass , Brian Belgodere , Sharath Pankanti
show 2 more authors
Noel Codella Patrick Watson
This is my paper · ORCID
classification cs.LGcs.AIcs.CV
keywords transferlearningsourcemodelselectionfine-tuningembeddingdivergencedatasetsizefeaturespacedistancesemanticrelationpredictionimageclassification
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 asks a practical question: when several pre-trained source models are available, which one should be fine-tuned for a new target task? It proposes an estimator, called P2L (Predict To Learn), that ranks sources using only the log-size of the source dataset plus a feature-space distance between source and target, computed with a single forward pass through one reference network. On the paper's 21-task validation, covering image classification and semantic-relation prediction, P2L selects the best source in 13 tasks and reaches 67.22% mean accuracy, versus 64.47% for the largest-dataset baseline and 64.86% for the least-divergent baseline. If these results hold, transfer-learning source selection no longer requires exhaustively fine-tuning every candidate.

What carries the argument

The load-bearing object is the estimator $E(t_i,s_j)$ itself. Each dataset is reduced to a single summary vector: the mean of penultimate-layer feature activations from a reference model (VGG16 for images, PCNN for relations), L1-normalized. $D(t_i,s_j)$ is then a divergence between those summaries, with KL and cityblock working well; $\log(|s_j|)$ encodes the empirically saturating benefit of larger sources; the $z$-scoring makes the two terms comparable; and $k$, tuned once on a small collection of tasks, balances them. Because the target needs only one forward pass through the reference network, the whole ranking costs seconds to minutes rather than a training run per candidate.

What would settle it

Run P2L on a set of source tasks that are entirely absent from the reference model's training data, exhaustively fine-tune all sources on the targets, and compare the predicted and true rankings; if the Spearman correlation drops to near zero, the distance signal is an artifact of the reference model rather than a general transfer property.

Watch

Extended reading notes

Core claim

The central claim is that transfer improvement $I(t_i,s_j)=P(M(t_i,s_j))-P(M(t_i,\phi))$ is ranked accurately by the embedding-divergence estimator $E(t_i,s_j)=z(\log(|s_j|))+k\,z(D(t_i,s_j))$, where $|s_j|$ is source dataset size, $D$ is a distance between dataset summaries in the penultimate layer of a fixed reference network, and $k$ is a single tuned scalar. The paper validates this by fine-tuning every source-target pair for 14 image tasks and 7 semantic-relation tasks, and reports Spearman correlations of 0.707 and 0.763 between predicted and true rankings in the two domains. It also reports that merging source datasets (ImageNet22K plus Places2) does not reliably beat the better individual source, which supports the view that transfer depends on compositional match, not just total data.

Load-bearing premise

The ranking collapses if the distance between source and target summaries in one fixed reference model's feature space is not a reliable proxy for fine-tuning improvement, especially since that reference model is trained on one of the candidate sources (ImageNet1K for images).

Editorial extensions

If this is right

  • A cloud service can recommend a fine-tuning source for a new customer dataset in about a minute per thousand images rather than training every candidate model.
  • The largest-dataset heuristic is not safe: across 21 tasks P2L outperforms it in mean accuracy and in the number of times the best source is selected.
  • The same estimator and the same tuned $k$ transfer across domains, working for both image classification and semantic-relation prediction.
  • The best source is usually near the top of P2L's ranking: the paper reports that finding it required at most three P2L picks for all 21 tasks, versus six for the largest-dataset method and five for the least-divergent method.

Reading between the lines

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

  • A natural extension is to pre-compute summary vectors for every model in a model zoo, turning source selection into a nearest-neighbor lookup in feature-vector space; the paper's setup already supplies the once-per-source computation.
  • Because the distance term is measured through one reference network, the method's reliability may depend on that reference; a testable variant would average the ranking over several reference networks trained on different data before trusting the recommendation.
  • The merged-dataset result suggests a counterintuitive practical rule: when expanding training data, adding examples from unrelated categories can reduce transfer value even as dataset size grows; this is an inference beyond the paper's experiments, but consistent with its size-versus-divergence trade-off.
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

3 major / 6 minor

Summary. The paper proposes P2L (Predict To Learn), an estimator E(t_i, s_j) = z(log(|s_j|)) + k·z(D(t_i, s_j)) that ranks candidate source models for transfer learning to a target task, using only the source dataset size and a feature-space distance between target and source datasets. The distance is computed by passing each dataset through a fixed reference model (VGG16 pretrained on ImageNet1K for images, a PCNN for semantic relations) and comparing the mean penultimate-layer activations. The parameter k is tuned on 9 ImageNet22K-derived tasks and then evaluated on 14 image tasks and 7 semantic-relation tasks, reporting Spearman correlations of 0.703 and 0.763 and an average accuracy of 67.22% versus 64.47% for the largest-dataset baseline and 64.86% for the least-divergent baseline (Table 1). The paper also presents an experiment on merged source datasets showing that a larger merged dataset does not always transfer better than ImageNet1K.

Significance. If the proposed estimator is a reliable predictor of transferability, it offers a practical and inexpensive way to select among pre-trained source models, which is of direct use for cloud-based training services. The held-out evaluation design—tuning k on one set of tasks and evaluating on disjoint image and relation tasks—is a genuine strength, as is the cross-domain transfer of k from images to relations. The merged-dataset experiment provides a useful counterexample to the assumption that larger sources are always better. However, the evidence is weakened by the possible confound in the distance measure, the lack of statistical significance reporting, and some presentation issues that affect the comparability of the headline numbers.

major comments (3)
  1. [Section 3.2 and Section 4.1] The distance D(t_i, s_j) is computed in the feature space of VGG16 pretrained on ImageNet1K, which is itself one of the 17 candidate source datasets listed in Section 4.1. Several target datasets (CUBS, Cars, Oxford, Wikiart) contain object categories that overlap ImageNet1K's class distribution. The penultimate-layer features of VGG16 are therefore optimized to separate exactly the categories that distinguish these sources and targets, which can make D(t, ImageNet1K) artificially small and inflate P2L's ranking accuracy without measuring true transferability. The paper provides no control experiment with a reference model trained on a domain disjoint from the candidate sources and targets. Without such a control, the reported Spearman correlations (Table 2 and Figure 6A) may reflect this overlap rather than a general property of the feature space. Please add an experiment using a reference model trained on a disjoint dataset (or a randomly initialized network) to demonstrate that the method is not reliant on the source pool being matched to the reference model's training data.
  2. [Section 4.3, Tables 1-4] The comparisons in Table 1 rely on point estimates with no confidence intervals or significance tests. Per-target Spearman rho values in Table 2 range from 0.52 to 0.95, and with only 17 sources per target, the uncertainty in each correlation is substantial. For semantic relations, the reported gain of 71.79 versus 70.6 for the largest-dataset baseline is small and may be within noise. To support the claim that P2L outperforms the baselines, please report task-level error bars or confidence intervals, and where meaningful, paired statistical tests across the 21 tasks.
  3. [Abstract and Table 1] The abstract states that P2L was validated 'across 95 tasks' and that the largest-dataset baseline selected the best model in only 55 cases, whereas the body evaluates 21 tasks and reports that P2L picked the best model in 13 cases; this discrepancy must be resolved. Additionally, Table 1 labels all values as 'Mean Top-1 Accuracy' even though the semantic-relations results are area under the PR curve, and the 'Average over 21 tasks' row averages raw values from two different metrics, which is not meaningful. Please report the two domains separately with their appropriate metrics and either remove the combined average or justify it explicitly.
minor comments (6)
  1. [Section 3.1 and Section 4.3.1] The paper says the estimator form was chosen 'based on extensive experimentation' but does not report the tuned value of k or the range over which it was searched; reporting these would improve reproducibility.
  2. [Section 4.1] The description of how target datasets are created from ImageNet22K partitions is ambiguous: 'One-tenth of the third partition was used to create a transfer learning target' does not specify the resulting number of training images per target task. Please clarify the exact construction procedure.
  3. [Section 5] There are typos in the Future Work section, including 'oc more enriched relationships' and 'may be provide'; the manuscript would benefit from a careful proofreading pass.
  4. [Section 4.3.3] The merged-dataset experiment with LC and ImageNet1K is interesting but is not explicitly connected to the P2L evaluation; state how this observation supports the main claim about the importance of both size and divergence.
  5. [Figure 5] The caption for Figure 5 is incomplete ('Log10(Size) of sourcesTv'); please provide a full caption describing the contents.
  6. [Table 3] The column header 'Least Divergent' refers to baseline B5, which uses the distance D; clarify this definition in the table caption or in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the P2L estimator is tuned on a disjoint 9-task set and evaluated on held-out image and relation tasks; the ImageNet reference-model overlap is a confound rather than a definitional reduction.

full rationale

The paper's estimator E(t_i,s_j)=z(log|s_j|)+k z(D(t_i,s_j)) (Eq. 2) is not fitted to the evaluation targets. Section 4.1 states that k is tuned on 9 ImageNet22K target tasks (St,Tt) and that 'The k value thus generated was used for evaluation on the workloads in the 14 target tasks (referred to above as Tv), as well as 7 tasks for Semantic Relations.' The image targets (CUBS, Cars, Sketches, Wikiart, Oxford, and the Visual Decathlon datasets) are distinct from those 9 tuning splits, and the relation targets are a separate domain. The ground-truth improvement I(t_i,s_j) is defined independently by fine-tuning (Eq. 1), while D(t_i,s_j) is computed from feature vectors of a reference network without using target-model accuracy. No equation defines E in terms of I, and no fitted parameter is renamed as a prediction on the held-out tasks. The only self-citations, [11] (related work on automatic labeling) and [14] (dataset source by a co-author), are descriptive and not load-bearing for the central estimator. The overlap of the VGG16 reference model's ImageNet1K pretraining with one candidate source and some targets is a genuine empirical confound that a strict evaluation should control for, but it does not make the reported ranking success equivalent to the input by construction; it is a benchmark-validity concern, not circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small number of modeling choices: the functional form of the estimator, the learned balancing weight k, the distance metric, and the reference model. The most consequential assumption is that the reference model's feature space is a valid proxy for transferability, an assumption validated only empirically on a modest number of tasks.

free parameters (2)
  • k = not reported in text
    Balancing weight between log-size and divergence in Equation 2, learned on the 9-task tuning set (St,Tt) to maximize Spearman correlation; then fixed for all evaluation tasks.
  • distance metric D = not explicitly reported; text suggests KL-divergence and cityblock work well, and Section 4.3.2 uses KLD
    Choice among KL, JSD, Chi-square, Euclidean, and cityblock, selected based on tuning-set performance in Figure 1; a discrete model choice affecting all predictions.
assumptions (4)
  • domain assumption The reference model (VGG16 for images, PCNN for relations) provides a feature space in which dataset distance correlates with transfer improvement.
    Introduced in Section 3.1 as the basis for D(t,s); validated only empirically on the tuning tasks, with no theoretical justification.
  • ad hoc to paper The estimator E(t,s) = z(log|s|) + k z(D(t,s)) is the correct functional form for predicting transfer improvement.
    Proposed in Section 3.1 based on 'extensive experimentation' and observed logarithmic size effect; not derived from first principles.
  • domain assumption A single learned weight k, fit on 9 ImageNet22K target tasks, generalizes to all image and semantic relation targets.
    Section 4.1 states the same k value is used for both domains and all evaluation targets.
  • domain assumption The evaluation datasets are representative of real transfer learning tasks.
    The paper uses common benchmarks plus a proprietary cloud API dataset; the latter is unavailable and its representativeness is not independently verifiable (Section 4.3.3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of P2L: Predicting Transfer Learning for Images and Semantic Relations." pith.science (2026). https://pith.science/paper/ZAMXNMWC

@misc{pith2026190807630,
  author       = {Pith},
  title        = {Pith review of: P2L: Predicting Transfer Learning for Images and Semantic Relations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZAMXNMWC}},
  note         = {Machine review of arXiv:1908.07630}
}
read the original abstract

Transfer learning enhances learning across tasks, by leveraging previously learned representations -- if they are properly chosen. We describe an efficient method to accurately estimate the appropriateness of a previously trained model for use in a new learning task. We use this measure, which we call "Predict To Learn" ("P2L"), in the two very different domains of images and semantic relations, where it predicts, from a set of "source" models, the one model most likely to produce effective transfer for training a given "target" model. We validate our approach thoroughly, by assembling a collection of candidate source models, then fine-tuning each candidate to perform each of a collection of target tasks, and finally measuring how well transfer has been enhanced. Across 95 tasks within multiple domains (images classification and semantic relations), the P2L approach was able to select the best transfer learning model on average, while the heuristic of choosing model trained with the largest data set selected the best model in only 55 cases. These results suggest that P2L captures important information in common between source and target tasks, and that this shared informational structure contributes to successful transfer learning more than simple data size.

Figures

Figures reproduced from arXiv: 1908.07630 by the authors.

Figure 1
Figure 1. Relationship of correlation metric (Spearman [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Deep Learning Pipeline for Images. Knowledge [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Deep Learning Pipeline for Relation Prediction. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Relationship of performance of the target model to size of the source dataset ( [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Log10(Size) of sources Tv these three methods, across images and relations. In the case of images, we present a comparison against ImageNet1K also in [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Number of attempts needed to get to best result for [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Accuracy: ImageNet1K vs. Combined dataset [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 34 canonical work pages

  1. [1]

    Zitnick K

    L. Zitnick K. He A. Torralba, K. Murphy. Visual domain decathlon. In PASCAL in Detail Workshop Challenge, CVPR , July 26th, Honolulu, Hawaii, USA. IEEE, 2017

    work page 2017

  2. [2]

    Task2vec: Task embedding for meta learning

    Rahul Tewari Avinash Ravichandran Subhransu Maji Charless Fowlkes Stefano Soatto Pietro Perona Alessandro Achille1, Michael Lam. Task2vec: Task embedding for meta learning. arXiv preprint arXiv:1902.03545, 2019

    arXiv 1902

  3. [3]

    Dbpedia: A nucleus for a web of open data

    Sören Auer, Christian Bizer, Georgi Kobilarov, Jens Lehmann, and Zachary Ives. Dbpedia: A nucleus for a web of open data. In In 6th Int’l Semantic Web Confer- ence, Busan, Korea, pages 11–15. Springer, 2007

    work page 2007

  4. [4]

    Factors of Transferability for a Generic ConvNet Representation

    Hossein Azizpour, Ali Sharif Razavian, Josephine Sul- livan, Atsuto Maki, and Stefan Carlsson. Factors of Transferability for a Generic ConvNet Representation. In IEEE Transactions on Pattern Analysis and Machine Intelligence, volume 38, pages 1790–1802, 2016

    work page 2016

  5. [5]

    J. Xiao A. Torralba B. Zhou, A. Lapedriza and A. Oliva. Learning deep features for scene recognition using places database. Proc. Advances Neural Inf. Process, 2014

    work page 2014

  6. [6]

    Relation adaptation: learning to extract novel relations with minimum supervision

    Danushka Bollegala, Yutaka Matsuo, and Mitsuru Ishizuka. Relation adaptation: learning to extract novel relations with minimum supervision. In Proceedings of the Twenty-Second international joint conference on Artificial Intelligence-Volume Volume Three, pages 2205–2210. AAAI Press, 2011

    work page

  7. [7]

    Stochastic gradient tricks

    Léon Bottou. Stochastic gradient tricks. In Gré- goire Montavon, Genevieve B. Orr, and Klaus-Robert Müller, editors, Neural Networks, Tricks of the Trade, Reloaded, Lecture Notes in Computer Science (LNCS 7700), pages 430–445. Springer, 2012

    work page 2012

  8. [8]

    Welinder P

    P. Welinder P. Perona C. Wah, S. Branson and S. Belongie. The caltech-ucsd birds-200-2011 dataset. 2011

    work page 2011

Show all 40 references
  1. [9]

    Boosting for transfer learning

    Wenyuan Dai, Qiang Yang, Gui-Rong Xue, and Yong Yu. Boosting for transfer learning. Intl Conf on Ma- chine Learning, pages 193–200, 2007

    2007

  2. [10]

    Document Image Classification with Intra-Domain Transfer Learning and Stacked Generalization of Deep Convolutional Neural Networks

    Arindam Das, Saikat Roy, and Ujjwal Bhattacharya. Document Image Classification with Intra-Domain Transfer Learning and Stacked Generalization of Deep Convolutional Neural Networks. 2018

    2018

  3. [11]

    Automatic labeling of data for transfer learning

    Parijat Dube, Bishwaranjan Bhattacharjee, Siyu Huo, Patrick Watson, Brian Belgodere, and John R Kender. Automatic labeling of data for transfer learning. nature, 192255:241, 2019

    2019

  4. [12]

    Model-agnostic meta-learning for fast adaptation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adaptation of deep networks. CoRR, abs/1703.03400, 2017

    2017 arXiv

  5. [13]

    Borrowing treasures from the wealthy: Deep transfer learning through selective joint fine-tuning

    Weifeng Ge and Yizhou Yu. Borrowing treasures from the wealthy: Deep transfer learning through selective joint fine-tuning. In Proc. IEEE Conference on Com- puter Vision and Pattern Recognition, Honolulu, HI , volume 6, 2017

    2017

  6. [14]

    A Dataset for Web- scale Knowledge Base Population

    Michael Glass and Alfio Gliozzo. A Dataset for Web- scale Knowledge Base Population. In Proceedings of the 15th Extended Semantic Web Conference, 2018

    2018

  7. [15]

    Diamos, Heewoo Jun, Hassan Kianinejad, Md

    Joel Hestness, Sharan Narang, Newsha Ardalani, Gre- gory F. Diamos, Heewoo Jun, Hassan Kianinejad, Md. Mostofa Ali Patwary, Yang Yang, and Yanqi Zhou. Deep learning scaling is predictable, empiri- cally. CoRR, abs/1712.00409, 2017

    2017 arXiv

  8. [16]

    Socher L.-J

    R. Socher L.-J. Li K. Li J. Deng, W. Dong and L. Fei-Fei. Imagenet: A large-scale hierarchical image database. IEEE Computer Vision and Pattern Recogni- tion (CVPR), 2009

    2009

  9. [17]

    J. Deng J. Krause, M. Stark and L. Fei-Fei. 3D object representations for fine-grained categorization. In In ICCV Workshop on Workshop on 3D Representation and Recognition, 2013

    2013

  10. [18]

    Multi-task transfer learning for weakly- supervised relation extraction

    Jing Jiang. Multi-task transfer learning for weakly- supervised relation extraction. In 4th International Joint Conference on Natural Language Processing. As- sociation for Computational Linguistics, 2009

    2009

  11. [19]

    The effects of training set size for performance of support vector machines and decision trees

    Taskin Kavzoglu and Ismail Colkesen. The effects of training set size for performance of support vector machines and decision trees. Symposium on Spatial Accuracy Assessment in Natural Resources, 2012

    2012

  12. [20]

    Metalearning: a survey of trends and technologies

    Christiane Lemke, Marcin Budka, and Bogdan Gabrys. Metalearning: a survey of trends and technologies. Ar- tificial Intelligence Review, 44(1):117–130, Jun 2015

    2015

  13. [21]

    Label efficient learning of transferable representations across domains and tasks

    Zelun Luo, Yuliang Zou, Judy Hoffman, and Li Fei-Fei. Label efficient learning of transferable representations across domains and tasks. In NIPS, 2017

    2017

  14. [22]

    J. Hays M. Eitz and M. Alexa. How do humans sketch objects? ACM Transactions on Graphics, 31(4):44–1, 2012

    2012

  15. [23]

    A Two-stage Approach for Extend- ing Event Detection to New Types via Neural Net- works

    Thien Huu Nguyen, Lisheng Fu, Kyunghyun Cho, and Ralph Grishman. A Two-stage Approach for Extend- ing Event Detection to New Types via Neural Net- works. ACL Representation Learning for NLP Work- shop, 2016

    2016

  16. [24]

    Nilsback and A

    M.-E. Nilsback and A. Zisserman. Automated flower classification over a large number of classes. In In Indian Conference on Computer Vision, Graphics & Image Processing, 2008

    2008

  17. [25]

    Berg Li Fei-Fei Olga Russakovsky, Jia Deng

    Hao Su Jonathan Krause Sanjeev Satheesh Sean Ma Zhiheng Huang Andrej Karpathy Aditya Khosla Michael Bernstein Alexander C. Berg Li Fei-Fei Olga Russakovsky, Jia Deng. Imagenet large scale visual recognition challenge. International Journal of Computer Vision, 2015

    2015

  18. [26]

    A Survey on Transfer Learning

    Sinno Jialin Pan and Qiang Yang. A Survey on Transfer Learning. IEEE Transactions on knowledge and data engineering, 22(10):1–15, 2010

    2010

  19. [27]

    Learning to learn, from transfer learning to domain adaptation: A unify- ing perspective

    Novi Patricia and Barbara Caputo. Learning to learn, from transfer learning to domain adaptation: A unify- ing perspective. Conference on Computer Vision and Pattern Recognition, 2014

    2014

  20. [28]

    CNN features off-the- shelf: an astounding baseline for recognition

    Ali Sharif Razavian, Hossein Azizpour, Josephine Sul- livan, and Stefan Carlsson. CNN features off-the- shelf: an astounding baseline for recognition. CoRR, abs/1403.6382, 2014

    2014 arXiv

  21. [29]

    To transfer or not to transfer

    Michael T Rosenstein, Zvika Marx, Leslie Pack Kael- bling, and Thomas G Dietterich. To transfer or not to transfer. In NIPS 2005 workshop on transfer learning, 2005

    2005

  22. [30]

    Learning to select data for transfer learning with bayesian optimization

    Sebastian Ruder and Barbara Plank. Learning to select data for transfer learning with bayesian optimization. CoRR, abs/1707.05246, 2017

    2017 arXiv

  23. [31]

    Saleh and A

    B. Saleh and A. Elgammal. Large-scale classification of fineart paintings: Learning the right metric on the right feature. 2015

    2015

  24. [32]

    Manning, and Andrew Y

    Richard Socher, Milind Ganjoo, Hamsa Sridhar, Osbert Bastani, Christopher D. Manning, and Andrew Y . Ng. Zero-Shot Learning Through Cross-Modal Transfer. pages 935–943, 2013

    2013

  25. [33]

    David Sussillo and L. Abbott. Transferring learning from external to internal weights in Echo-State net- works with sparse connectivity. PLoS ONE, 2012

    2012

  26. [34]

    Transfer learning

    Lisa Torrey and Jude Shavlik. Transfer learning. In Handbook of research on machine learning applica- tions and trends: algorithms, methods, and techniques, pages 242–264. IGI Global, 2010

    2010

  27. [35]

    Learning to transfer

    Ying Wei, Yu Zhang, and Qiang Yang. Learning to transfer. CoRR, abs/1708.05629, 2017

    2017 arXiv

  28. [36]

    Khoshgoftaar, and DingDing Wang

    Karl Weiss, Taghi M. Khoshgoftaar, and DingDing Wang. A survey of transfer learning. Journal of Big Data, 3(1):9, May 2016

    2016

  29. [37]

    Transfer learning via learning to transfer

    Wei Ying, Yu Zhang, Junzhou Huang, and Qiang Yang. Transfer learning via learning to transfer. In Interna- tional Conference on Machine Learning, pages 5072– 5081, 2018

    2018

  30. [38]

    How transferable are features in deep neural networks? In NIPS, 2014

    Jason Yosinski, Jeff Clune, Yoshua Bengio, and Hod Lipson. How transferable are features in deep neural networks? In NIPS, 2014

    2014

  31. [39]

    Shen, Leonidas J

    Amir Roshan Zamir, Alexander Sax, William B. Shen, Leonidas J. Guibas, Jitendra Malik, and Silvio Savarese. Taskonomy: Disentangling task transfer learning. CoRR, abs/1804.08328, 2018

    2018 arXiv

  32. [40]

    Distant supervision for relation extraction via piece- wise convolutional neural networks

    Daojian Zeng, Kang Liu, Yubo Chen, and Jun Zhao. Distant supervision for relation extraction via piece- wise convolutional neural networks. In EMNLP, pages 1753–1762, 2015

    2015

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.