Pith. sign in

REVIEW 3 major objections 8 minor 31 references

Eidetic Learning: an Efficient and Provable Solution to Catastrophic Forgetting

T0 review · 3 major / 8 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A network trained with Eidetic Learning cannot forget: freezing a task's important neurons and deleting synapses from recycled neurons into them makes its output on that task exactly immutable.

desk verdict A solid task-incremental continual learning method built on prune-and-freeze, with a sound core proof for linear and convolutional layers; the BN exactness claim is under-specified and the abstract overreaches, but the mechanism deserves serious engagement. read the letter →

arxiv 2502.09500 v2 pith:454U624U submitted 2025-02-13 cs.LG

Nicholas Dronen , Randall Balestriero This is my paper · ORCID
classification cs.LG
keywords catastrophicforgettingcontinuallearningiterativepruningneuronfreezingtask-incrementalmixtureofexpertsbatchnormalization
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 sets out to make catastrophic forgetting impossible by construction rather than less likely. Its proposed training procedure, Eidetic Learning, trains each task, freezes the neurons that task needs, deletes every synapse leading from the remaining 'recycled' neurons into those frozen neurons, and then reuses the recycled neurons for the next task. The claim is that any network trained this way—an EideticNet—produces exactly the same outputs on every previously learned task after later tasks are trained, with no rehearsal or replay. If true, continual learning in overparameterized networks becomes a capacity-budgeting problem: as long as each new task fits in the spare neurons, old skills are retained exactly. The paper also shows how a learned task classifier lets the network route new examples without being told which task they come from.

What carries the argument

The object that carries the argument is the pair of conditions (I) Persistence and (II) Resistance, stated in Section 3 as sufficient for exact retention of $f_{\theta_{t_j}}(x_i)$. The implementing mechanism is iterative structured pruning with a nested sparsity pattern: a per-task neuron mask selects the minimal set $N_{t_i}$; frozen neurons $F$ accumulate; neurons in $R = N \setminus F$ are reinitialized and their synapses into $F$ are permanently zeroed. For batch-normalization layers, the learned scale and shift are pruned with the layer mask, and the running statistics of frozen channels must be kept in evaluation mode and not updated during later tasks.

What would settle it

On a model with batch normalization, train an EideticNet on task 1, freeze its important neurons and their batch-normalization running statistics, then train task 2 in ordinary training mode; recompute task 1's outputs. If the frozen channels' running mean or variance moved, or if any task-1 test input has a different output before and after task 2, exact immutability is refuted for modern normalization-layer networks.

Watch

Extended reading notes

Core claim

The paper's central claim is that two conditions—persistence (neurons important to task $t_i$ are unchanged during later tasks) and resistance (no signal flows from non-important neurons into those important neurons)—are together sufficient to make $f_{\theta_{t_j}}(x_i)$ identical to $f_{\theta_{t_i}}(x_i)$ for every input $x_i$ and every later task $t_j$. Eidetic Learning implements these conditions by iterative structured pruning: train a task, prune the smallest set of neurons whose removal keeps training accuracy above a threshold, freeze the survivors, delete the synapses from pruned to frozen neurons, reinitialize the pruned neurons, and train the next task. The unusual architectural choice is nesting: later tasks can use earlier tasks' frozen features through remaining synapses, while the deleted synapses guarantee that recycled neurons cannot affect frozen ones. If the implementation really holds batch-normalization running statistics fixed per frozen channel, no later training step can alter any earlier task's function.

Load-bearing premise

The proof of exact retention assumes that a frozen channel's batch-normalization running statistics can be held bit-for-bit at their end-of-task values while other channels train in standard mode, but standard frameworks update all channels' statistics during training and the paper specifies no per-channel freezing mechanism.

Editorial extensions

If this is right

  • If the two conditions are met, the network's output on every past task is immutable, so forgetting is impossible by construction rather than reduced in probability.
  • Because important neurons are frozen and incoming synapses from recycled neurons are deleted, later tasks can reuse early-task features without disturbing them; only forward transfer is supported.
  • The method needs no rehearsal or replay and adds only a constant number of hyperparameters per network, with time and space complexity linear in parameter count.
  • A learned task classifier on the penultimate hidden states removes the need for a task ID at inference, at some measured cost in per-class accuracy in the reported CIFAR-100 experiments.
  • The same guarantee extends to convolutional, residual, and normalization layers, and to fine-tuning after pre-training, as long as frozen batch-normalization statistics are held fixed.

Reading between the lines

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

  • Editorial inference: if exact immutability holds, catastrophic forgetting is reframed as a capacity-allocation problem: an EideticNet is correct whenever each new task fits in the remaining recycled capacity, making 'excess capacity' a measurable resource.
  • Editorial inference: the method's explicit assumption of task-incremental learning suggests a natural test—overlapping classes would require a different routing or multi-head arbitration mechanism, since the learned task classifier is supervised.
  • Editorial inference: the backward-transfer extension described in the paper would let later-learned features improve earlier tasks, but it requires re-training the earlier task on its original data when the data distribution is non-stationary, which is an implicit replay requirement.
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 / 8 minor

Summary. The paper proposes Eidetic Learning, a continual-learning method that trains each task, identifies important neurons by iterative structured pruning, freezes those neurons, and prunes all synapses from pruned/recycled neurons into the frozen neurons before training the next task. The authors argue that these two operations, Persistence and Resistance, are sufficient to keep the network's output for every previous task exactly unchanged, so catastrophic forgetting is impossible by construction. The method is implemented for linear, convolutional, batch-normalization, and residual layers, and it includes a learned task classifier that routes new instances to the per-task classifier without explicit task IDs. Experiments on Permuted MNIST, Sequential CIFAR100, and Sequential Imagenette show flat test accuracy for earlier tasks and competitive per-task accuracy, with code released.

Significance. If the construction is implemented exactly as described, the paper offers a clean and practically useful sufficient condition for avoiding catastrophic forgetting: freezing task-important neurons and deleting incoming synapses from recycled neurons makes the representation of an earlier task invariant to later training. The presented conditions are transparent, the empirical curves are consistent with the claimed invariance, the method requires no replay, and the open-source release is a concrete strength. The paper also contributes an honest comparison against related pruning-based continual-learning methods and reports the cost of learned task routing. The main limitations are that the BatchNorm preservation mechanism is not specified at the per-channel level, which is necessary for the exactness proof in modern networks, and several claims about 'smallest' sets and end-to-end guarantees go beyond what is established.

major comments (3)
  1. [Section 3.1, Eq. (2)] The exactness guarantee for BatchNorm layers is not established. The text says that for previous tasks the learnable parameters remain in evaluation mode and that their internal running statistics are not updated, but a standard BatchNorm module has a single module-level training flag and updates running_mean and running_var for every channel on each training forward pass. If the whole BN module is kept in train mode, the frozen channels' statistics drift; if it is kept in eval mode, the recycled channels can no longer learn. The paper does not specify a per-channel BatchNorm implementation that freezes only the statistics of important neurons while allowing recycled channels to update. Without such a mechanism, the values of mu and sigma entering frozen channels change during later-task training, so the output of the frozen subnetwork is no longer exactly preserved and the proof's premise fails for the ResNet experiments. Please specify the exact per-channel masking procedure or restrict the formal guarantee to architectures without normalization layers.
  2. [Section 3, Step I and Section 2] The algorithm is described as finding 'the smallest set of neurons necessary to perform task ti', and the introduction and Listing 1 repeat the claim that the framework selects a 'minimal' subset of neurons. However, the actual iterative-pruning procedure uses a fixed pruning step size and a stopping threshold on training-set accuracy; the paper gives no proof, and no plausible argument, that this procedure produces a minimal set of neurons. This is load-bearing for the capacity-efficiency and interpretability claims, because the method's stated benefit is that it accounts for the 'incremental amount of capacity required'. Please either provide a formal treatment of minimality or replace 'smallest' and 'minimal' with language describing an approximately-minimal set determined by the pruning hyperparameters.
  3. [Abstract and Section 4, Tables 7 and 8] The no-forgetting guarantee applies to the per-task representation under oracle task routing, while the abstract claims that an EideticNet 'automatically routes new instances without auxiliary task information' and thereby 'provably solves catastrophic forgetting'. The task classifier is a separate learned component that is not covered by the Persistence/Resistance argument, and the experiments show that it can make substantial routing errors: for the small MLP on CIFAR10-5, oracle routing achieves 80.60% on class 0 with l1 pruning while the learned router achieves 63.40%, and for ResNet50 the average per-class drop is 1.8 percentage points. The manuscript should explicitly state that the exactness guarantee is conditional on correct task routing and should specify when and how the task classifier is trained or updated across the task sequence, since an incrementally updated classifier could itself suffer from forgetting.
minor comments (8)
  1. [Abstract] The sentence 'we believe they can be benefit practitioners and theorists alike' contains a grammatical error ('be benefit') that should be corrected.
  2. [Section 1] The phrase 'seteting the targets yit of each specific task t to t' has a typo: 'seteting' should be 'setting'.
  3. [Section 2] The related-work section uses 'CNLP' in Table 1 but 'CLNP' in the text; the abbreviation should be consistent.
  4. [Section 3.1] The sentence 'we also to (i) make the beta, gamma of the previous tasks stay in evaluation mode...' is incomplete; the missing verb should be supplied. This is especially important because this sentence is the entire description of the BatchNorm preservation mechanism.
  5. [Section 3.1] The text promises 'an argument for why making self-attention layers of Transformers immune to catastrophic forgetting is challenging if not impossible', but no such argument appears in the manuscript. Either add the argument or remove the promise.
  6. [Section 5] The text contains several typos, including 'F orward transfer' with an extra space and 'Future work may extending Eidetic Learning' with an ungrammatical verb form.
  7. [Appendix, Section 6.1] The appendix still contains the template sentence 'Authors may wish to optionally include extra information...' which should be removed from a submitted manuscript.
  8. [Table 3] The table reports standard deviations for the authors' method but not for the cited baselines; please clarify whether the baseline values are taken directly from the original papers and whether the comparison is on identical train/test splits.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the forgetting guarantee follows directly from freezing and pruning, with no fitted value renamed as a prediction.

full rationale

The paper's central claim is that Persistence (freezing neurons important to each task) and Resistance (pruning synapses from recycled neurons into frozen neurons) make the network output for every previous task exactly unchanged, so forgetting is impossible by construction. This is a direct constructive proof rather than a circular reduction: the method is designed to satisfy two stated conditions, and the implication from those conditions to output immutability is a formal consequence of the layer equations given in Section 3.1, not an equivalence between a fitted input and a predicted output. The task classifier used for routing is a separate supervised component evaluated against oracle routing, and its accuracy is not used to establish the forgetting guarantee. No load-bearing self-citations were found; the sparsity pattern is attributed to prior work by Golkar et al. and independently discovered, and the paper does not rely on a uniqueness theorem or ansatz smuggled in via citation. The reviewer's concern about batch-normalization running statistics is a correctness or implementation gap for the exactness guarantee, but it is not circularity: it does not make the derivation equivalent to its inputs. The experimental comparisons are against external baselines and do not reduce to fitted parameters. Overall, no pattern of self-definitional reasoning, fitted-input prediction, or citation-backed circularity is present.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method relies on standard network operations plus three domain assumptions: spare capacity, per-channel BN freezing, and a usable task router. No new physical or statistical entities are introduced.

free parameters (6)
  • pruning step size = 1% or 5% of neurons per layer per iteration
    Chosen by held-out validation (Table 2); controls how fast neurons are pruned during iterative pruning.
  • stop threshold = 0.1% to 3% below max training accuracy
    Chosen by held-out validation (Table 2); stops pruning before training accuracy degrades.
  • dropout probability = 0.1 for Permuted MNIST
    Selected in hyperparameter search over {0, 0.03, 0.10}; used only for the MLP experiment.
  • learning rate reduction flag = True for Permuted MNIST
    Whether to halve the learning rate before pruning; selected in hyperparameter search.
  • early stopping patience = 10 epochs
    Fixed for all experiments; used to detect convergence during training and recovery.
  • max recovery epochs = 5 for PMNIST, 10 for deep networks
    Maximum epochs allowed to retrain after a pruning round; listed in Section 4.
assumptions (4)
  • standard math A feed-forward network made of linear layers and pointwise nonlinearities computes a function completely determined by its active neurons when all incoming synapses from pruned neurons are zero.
    Used in Section 3 to argue that pruning and freezing make a task's function immutable.
  • domain assumption The network has excess capacity for every task, meaning a subnetwork can reach near-full training accuracy and leave enough recycled neurons for later tasks.
    Stated in Section 5: 'Training one to good effect requires that excess capacity exists in the ANN.' Without it, no neurons can be recycled and the method cannot train new tasks.
  • domain assumption BatchNorm running statistics for frozen channels can be preserved exactly while new tasks train, even though standard BN updates all channels in training mode.
    Section 3.1 claims previous tasks' BN is kept in evaluation mode, but per-channel freezing of running statistics is not standard and not fully specified.
  • domain assumption The learned task classifier routes new instances to the correct task head well enough that routing errors remain a minor factor in overall accuracy.
    Section 4 reports a 1.8% average drop from oracle routing on CIFAR100-5; the no-task-ID claim depends on this classifier.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Eidetic Learning: an Efficient and Provable Solution to Catastrophic Forgetting." pith.science (2026). https://pith.science/paper/454U624U

@misc{pith2026250209500,
  author       = {Pith},
  title        = {Pith review of: Eidetic Learning: an Efficient and Provable Solution to Catastrophic Forgetting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/454U624U}},
  note         = {Machine review of arXiv:2502.09500}
}
read the original abstract

Catastrophic forgetting -- the phenomenon of a neural network learning a task t1 and losing the ability to perform it after being trained on some other task t2 -- is a long-standing problem for neural networks [McCloskey and Cohen, 1989]. We present a method, Eidetic Learning, that provably solves catastrophic forgetting. A network trained with Eidetic Learning -- here, an EideticNet -- requires no rehearsal or replay. We consider successive discrete tasks and show how at inference time an EideticNet automatically routes new instances without auxiliary task information. An EideticNet bears a family resemblance to the sparsely-gated Mixture-of-Experts layer Shazeer et al. [2016] in that network capacity is partitioned across tasks and the network itself performs data-conditional routing. An EideticNet is easy to implement and train, is efficient, and has time and space complexity linear in the number of parameters. The guarantee of our method holds for normalization layers of modern neural networks during both pre-training and fine-tuning. We show with a variety of network architectures and sets of tasks that EideticNets are immune to forgetting. While the practical benefits of EideticNets are substantial, we believe they can be benefit practitioners and theorists alike. The code for training EideticNets is available at https://github.com/amazon-science/eideticnet-training.

Figures

Figures reproduced from arXiv: 2502.09500 by the authors.

Figure 1
Figure 1. Accuracy of an MLP trained on 10 tasks of Permuted MNIST in a single run of our method. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Eidetic Learning eliminates forgetting by preserving important neurons, deleting unim [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. In Eidetic Training, the fraction of neurons pruned while training a task increases until [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Test set accuracy of ResNet50 trained on Sequential CIFAR100 with five tasks. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Consider a feed-forward ANN with layers ℓ, ℓ + 1 trained on some task ti . Omitting for convenience the non-linearity σ and the bias b, processing the input x ℓ−1 vector of all 1s entails the matrix-vector products Wℓx ℓ−1 and Wℓ+1x ℓ . We show them here as the composi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 20 canonical work pages

  1. [1]

    Superposition of many models into one

    Brian Cheung, Alexander Terekhov, Yubei Chen, Pulkit Agrawal, and Bruno Olshausen. Superposition of many models into one. Advances in neural information processing systems, 32, 2019

    work page 2019

  2. [2]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248--255. Ieee, 2009

    2009

  3. [3]

    Head2toe: Utilizing intermediate representations for better transfer learning

    Utku Evci, Vincent Dumoulin, Hugo Larochelle, and Michael C Mozer. Head2toe: Utilizing intermediate representations for better transfer learning. In International Conference on Machine Learning, pages 6009--6033. PMLR, 2022

    work page 2022

  4. [4]

    Continual learning via neural pruning

    Siavash Golkar, Michael Kagan, and Kyunghyun Cho. Continual learning via neural pruning. arXiv preprint arXiv:1903.04476, 2019

    arXiv 1903

  5. [5]

    Shels: Exclusive feature sets for novelty detection and continual learning without class boundaries

    Meghna Gummadi, David Kent, Jorge A Mendez, and Eric Eaton. Shels: Exclusive feature sets for novelty detection and continual learning without class boundaries. In Conference on Lifelong Learning Agents, pages 1065--1085. PMLR, 2022

    work page 2022

  6. [6]

    Learning both weights and connections for efficient neural network

    Song Han, Jeff Pool, John Tran, and William Dally. Learning both weights and connections for efficient neural network. Advances in neural information processing systems, 28, 2015

    2015

  7. [7]

    Imagenette: A smaller subset of 10 easily classified classes from imagenet, March 2019

    Jeremy Howard. Imagenette: A smaller subset of 10 easily classified classes from imagenet, March 2019. URL https://github.com/fastai/imagenette

    work page 2019

  8. [8]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning, pages 448--456. pmlr, 2015

    2015

Show all 31 references
  1. [9]

    Russian doll network: Learning nested networks for sample-adaptive dynamic inference

    Borui Jiang and Yadong Mu. Russian doll network: Learning nested networks for sample-adaptive dynamic inference. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 336--344, 2021

    2021

  2. [10]

    Continual learning with node-importance based adaptive group sparse regularization

    Sangwon Jung, Hongjoon Ahn, Sungmin Cha, and Taesup Moon. Continual learning with node-importance based adaptive group sparse regularization. Advances in neural information processing systems, 33: 0 3647--3658, 2020

    2020

  3. [11]

    Natural continual learning: success is a journey, not (just) a destination

    Ta-Chu Kao, Kristopher Jensen, Gido van de Ven, Alberto Bernacchia, and Guillaume Hennequin. Natural continual learning: success is a journey, not (just) a destination. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan, editors, Advances in Neural I...

    2021

  4. [12]

    Understanding catastrophic forgetting and remembering in continual learning with optimal relevance mapping

    Prakhar Kaushik, Adam Kortylewski, Alex Gain, and Alan Yuille. Understanding catastrophic forgetting and remembering in continual learning with optimal relevance mapping. In Fifth Workshop on Meta-Learning at the Conference on Neural Information Processing Systems, 2021

    2021

  5. [13]

    Nestednet: Learning nested sparse structures in deep neural networks

    Eunwoo Kim, Chanho Ahn, and Songhwai Oh. Nestednet: Learning nested sparse structures in deep neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

    2018

  6. [14]

    Overcoming catastrophic forgetting in neural networks

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Scien...

    2017

  7. [15]

    Packnet: Adding multiple tasks to a single network by iterative pruning

    Arun Mallya and Svetlana Lazebnik. Packnet: Adding multiple tasks to a single network by iterative pruning. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 7765--7773, 2018

    2018

  8. [16]

    Piggyback: Adapting a single network to multiple tasks by learning to mask weights

    Arun Mallya, Dillon Davis, and Svetlana Lazebnik. Piggyback: Adapting a single network to multiple tasks by learning to mask weights. In Proceedings of the European conference on computer vision (ECCV), pages 67--82, 2018

    2018

  9. [17]

    Class-incremental learning: survey and performance evaluation on image classification

    Marc Masana, Xialei Liu, Bart omiej Twardowski, Mikel Menta, Andrew D Bagdanov, and Joost Van De Weijer. Class-incremental learning: survey and performance evaluation on image classification. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (5): 0 5513--5533, 2022

    2022

  10. [18]

    Parallel distributed processing, volume 2: Explorations in the microstructure of cognition: Psychological and biological models, volume 2

    James L McClelland, David E Rumelhart, PDP Research Group, et al. Parallel distributed processing, volume 2: Explorations in the microstructure of cognition: Psychological and biological models, volume 2. MIT press, 1987

    1987

  11. [19]

    Catastrophic interference in connectionist networks: The sequential learning problem

    Michael McCloskey and Neal J Cohen. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, pages 109--165. Elsevier, 1989

    1989

  12. [20]

    Pruning convolutional neural networks for resource efficient inference

    Pavlo Molchanov, Stephen Tyree, Tero Karras, Timo Aila, and Jan Kautz. Pruning convolutional neural networks for resource efficient inference. In International Conference on Learning Representations, 2017. URL https://openreview.net/forum?id=SJGCiw5gl

    2017

  13. [21]

    Online structured laplace approximations for overcoming catastrophic forgetting

    Hippolyt Ritter, Aleksandar Botev, and David Barber. Online structured laplace approximations for overcoming catastrophic forgetting. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors, Advances in Neural Information Processing System...

    2018

  14. [22]

    Parallel distributed processing, volume 1: Explorations in the microstructure of cognition: Foundations

    David E Rumelhart, James L McClelland, PDP Research Group, et al. Parallel distributed processing, volume 1: Explorations in the microstructure of cognition: Foundations. The MIT press, 1986

    1986

  15. [23]

    Outrageously large neural networks: The sparsely-gated mixture-of-experts layer

    Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. In International Conference on Learning Representations, 2016

    2016

  16. [24]

    Dropout: a simple way to prevent neural networks from overfitting

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 15 0 (1): 0 1929--1958, 2014

    1929

  17. [25]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. Attention is all you need. In I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural In...

    2017

  18. [26]

    Continual learning through synaptic intelligence

    Friedemann Zenke, Ben Poole, and Surya Ganguli. Continual learning through synaptic intelligence. In Doina Precup and Yee Whye Teh, editors, Proceedings of the 34th International Conference on Machine Learning, volume 70 of Proceedings of Machine Learning Research, pages 3987-...

    2017

  19. [27]

    Understanding deep learning requires rethinking generalization

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. Understanding deep learning requires rethinking generalization. In International Conference on Learning Representations, 2017. URL https://openreview.net/forum?id=Sy8gdB9xx

    2017

  20. [28]

    Understanding deep learning (still) requires rethinking generalization

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. Understanding deep learning (still) requires rethinking generalization. Communications of the ACM, 64 0 (3): 0 107--115, 2021

    2021

  21. [29]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  22. [30]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  23. [31]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

    2024

Pith tools

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