The reviewed record of science sign in
Pith

arxiv: 1708.07747 · v2 · pith:MZV6RJ7N · submitted 2017-08-25 · cs.LG · cs.CV· stat.ML

Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

Han Xiao , Kashif Rasul , Roland Vollgraf This is my paper

Reviewed by Pith2026-05-11 05:17 UTCgrok-4.3pith:MZV6RJ7Nopen to challenge →

classification cs.LG cs.CVstat.ML
keywords Fashion-MNISTimage classification datasetMNIST replacementmachine learning benchmarksfashion product imagesgrayscale 28x28 imagesdrop-in dataset
0
0 comments X

The pith

Fashion-MNIST supplies a drop-in replacement for MNIST using 28x28 fashion images.

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

The paper releases Fashion-MNIST, a collection of 70,000 grayscale images showing fashion products from ten categories. It keeps exactly the same 28-by-28 size, single-channel format, and 60,000/10,000 train-test split as the original MNIST handwritten-digit set. The change in subject matter is meant to raise the difficulty of the classification task while preserving every detail of the evaluation protocol. Researchers can therefore swap the dataset into existing code and obtain more informative benchmark numbers without any other changes.

Core claim

Fashion-MNIST is a new dataset of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set contains 60,000 images and the test set contains 10,000 images. The dataset is constructed to function as a direct drop-in replacement for the original MNIST dataset, matching its image size, data format, and training-testing split structure exactly.

What carries the argument

The exact structural match to MNIST (image dimensions, grayscale format, and split sizes) applied to a new subject domain of clothing items.

If this is right

  • Existing benchmark code and leaderboards can be reused unchanged while testing on more varied visual content.
  • Performance gaps between models will more accurately reflect generalization beyond simple digit shapes.
  • New algorithms can be compared directly against prior work without needing to re-implement MNIST baselines.
  • The dataset remains freely downloadable and usable under the same conditions as the original MNIST.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Widespread adoption would discourage overfitting to the specific visual statistics of handwritten digits.
  • The format match could inspire similar replacements for other long-standing but overly simple benchmarks.
  • Developers of feature-extraction methods would need to handle intra-class variation in texture and shape that digits lack.

Load-bearing premise

That the fashion images will prove meaningfully harder for models yet still accessible enough that the community will switch to this dataset instead of continuing to use MNIST.

What would settle it

A broad survey of recent papers that shows most new algorithms still report results only on MNIST and not on Fashion-MNIST.

read the original abstract

We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at https://github.com/zalandoresearch/fashion-mnist

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript introduces Fashion-MNIST, a dataset of 70,000 28x28 grayscale images of fashion products across 10 categories (7,000 images per category), with a 60,000-image training set and 10,000-image test set. It is explicitly positioned as a direct drop-in replacement for the original MNIST dataset, matching it in image size, data format (IDX binary), and train/test split structure. The dataset is released publicly via the cited GitHub repository.

Significance. If adopted, the dataset offers a more challenging yet compatible benchmark for image classification algorithms, addressing MNIST's simplicity while preserving reproducibility and ease of use in existing pipelines. The public release, identical format, and clear specification of splits constitute a concrete contribution that enables immediate community use and more realistic model evaluations.

minor comments (2)
  1. [Abstract] Abstract: the phrasing 'comprising of' is nonstandard; 'consisting of' or 'comprising' would be clearer.
  2. [Dataset description] The manuscript would benefit from a short table or paragraph in the main text explicitly comparing the exact file formats and split sizes to MNIST to strengthen the drop-in claim.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript and for recommending acceptance. The review accurately captures the intent and contribution of Fashion-MNIST as a direct drop-in replacement for MNIST.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a dataset release note with no derivations, equations, predictions, fitted parameters, or theoretical claims. The central statement that Fashion-MNIST matches MNIST in size, format, and split structure is a direct description of the released data files themselves (publicly provided in the cited GitHub repository in identical IDX format). No load-bearing step reduces to a self-citation, ansatz, or input-by-construction; the format equivalence is verifiable externally from the dataset release without any internal loop.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a dataset introduction paper with no mathematical derivations, models, or theoretical claims, so the ledger contains no free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5399 in / 1089 out tokens · 52756 ms · 2026-05-11T05:17:28.596879+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 60 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Gradient-Free Continual Learning in Spiking Neural Networks via Inter-Spike Interval Regularization

    cs.NE 2026-04 unverdicted novelty 8.0

    ISI-CV derives a synaptic importance score from the regularity of neuron firing intervals to enable continual learning without gradients or forgetting on SNNs.

  2. Model Merging as Probabilistic Inference in Fine-Tuning Parameter Space

    cs.LG 2026-07 unverdicted novelty 7.0

    Model merging is cast as PoE inference with EBM experts, revealing Gaussian assumptions in prior work and proposing convergent Cauchy experts that improve empirical performance.

  3. Effective Covariance Dynamics in Solvable High-Dimensional GANs

    cs.LG 2026-06 unverdicted novelty 7.0

    Extends solvable GAN theory to structured latent covariances, proving convergence of stochastic training to ODEs and identifying a correlation-based signal-boosting mechanism for subspace recovery.

  4. Training-free Task Classification for Multi-Task Model Merging

    cs.LG 2026-06 conditional novelty 7.0

    SiM enables training-free routing in multi-task model merging by scoring test inputs via projection residuals onto SVD-based task manifolds precomputed from small support sets.

  5. Fast Computation of Free-Support Wasserstein Medians

    stat.CO 2026-06 unverdicted novelty 7.0

    Direct fixed-weight solver for free-support Wasserstein medians relocates atoms using OT barycentric projections and inverse-distance weights, achieving monotone descent on smoothed objectives with fewer subproblems t...

  6. GrapNet: A Programmable Dynamic-Architecture Neural Graph Substrate

    cs.LG 2026-06 unverdicted novelty 7.0

    GrapNet defines a graph-as-architecture neural substrate with node-owned child references and allocation vectors that supports structural edits and shows accuracy gains over MLPs in continual learning on Split Fashion...

  7. A Unified Zeroth-Order Approach for Decentralized Minimax Optimization

    math.OC 2026-06 unverdicted novelty 7.0

    ZOMA unifies hybrid zeroth-order estimators, bias corrections (GT/ED/EXTRA), and accelerations (STORM/PAGE/L2S) for decentralized nonconvex PL minimax optimization, claiming convergence rates matching centralized meth...

  8. End-to-End Optimization of Incoherent Imaging for Classification Under Detector-Limited Readout

    cs.CV 2026-06 unverdicted novelty 7.0

    End-to-end phase-mask optimization improves classification under constrained detector readout by increasing class separability but yields no benefit under full readout, where a conventional lens approaches the mutual-...

  9. Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors

    stat.ML 2026-06 unverdicted novelty 7.0

    Helmholtz-Hodge decomposition of score errors shows only the gradient component affects marginal Fokker-Planck dynamics in diffusion models, yielding an impossibility result for L2 error bounding divergences and a tra...

  10. Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors

    stat.ML 2026-06 unverdicted novelty 7.0

    Only the gradient component of score errors affects marginal distributions in diffusion models, so L2 error can be arbitrarily large with perfect match; this yields an impossibility result, a gradient-only KL bound, a...

  11. Compress then Merge: From Multiple LoRAs into One Low-Rank Adapter

    cs.LG 2026-06 unverdicted novelty 7.0

    CtM merges T LoRAs into one rank-r LoRA by computing shared r-dimensional subspaces from the LoRA weights, projecting adapters into r x r coordinates, and merging in that reduced space, outperforming merge-then-compre...

  12. Augmented Lagrangian Predictive Coding

    cs.LG 2026-05 unverdicted novelty 7.0

    PC-ALM uses dual ascent on an augmented Lagrangian to achieve exact backpropagation gradients via layer-local updates in linear networks and matching performance in nonlinear networks up to depth 128.

  13. Quantum State Preparation via Neural Network Encoding in Quantum Machine Learning

    quant-ph 2026-05 unverdicted novelty 7.0

    A neural network is trained to predict parameters of a fixed quantum circuit, enabling high-fidelity quantum state preparation from classical data in one inference step with up to 0.992 fidelity on unseen MNIST and Fa...

  14. ScaleMAP: Preserving Local Density and Neighborhood Structure in Low-Dimensional Embeddings

    cs.LG 2026-05 unverdicted novelty 7.0

    ScaleMAP is a dimensionality-reduction method that preserves both neighborhood structure and local density by scaling embedding displacements with original local radii, matching DensMAP on density while retaining UMAP...

  15. Complement Submodular Information Measures for Balanced and Robust Data Selection

    cs.LG 2026-05 unverdicted novelty 7.0

    Introduces complement-aware submodular functions (CSI) that preserve structure between subset and complement for improved robust data selection.

  16. Testing the Test: Score-Direction Instability in Class-Split Anomaly Detection

    cs.LG 2026-05 unverdicted novelty 7.0

    Class-split anomaly detection benchmarks can yield collapsing or inverted anomaly scores when the held-out class overlaps normal data in representation space, with neighborhood class leakage predicting this instabilit...

  17. Pointwise Generalization in Deep Neural Networks

    cs.LG 2026-05 unverdicted novelty 7.0

    Proposes pointwise Riemannian Dimension from feature eigenvalues to derive tighter, representation-aware generalization bounds for deep networks in the nonlinear regime.

  18. BESplit: Bias-Compensated Split Federated Learning with Evidential Aggregation

    cs.LG 2026-05 unverdicted novelty 7.0

    BESplit mitigates non-IID bias in split federated learning via evidential aggregation, bias-compensated client pairing, and dual-teacher distillation, outperforming prior methods on five benchmarks.

  19. PCDM: A Diffusion-Based Data Poisoning Attack Against Federated Learning Systems

    cs.CR 2026-05 unverdicted novelty 7.0

    PCDM uses a poisoning-oriented conditional diffusion model with an adjustable vector and jumping strategy to create stealthier and more effective poisoned data than GAN-based attacks against federated learning.

  20. Byzantine-Resilient Federated Learning via QUBO-Based Client Selection on Quantum Annealers

    cs.LG 2026-05 unverdicted novelty 7.0

    QUBO formulation on quantum annealers for joint client selection in federated learning, combined with a MultiSignal routing ensemble, yields higher Byzantine attack detection accuracy than MultiKrum on challenging att...

  21. Not All Symbols Are Equal: Importance-Aware Constellation Design for Semantic Communication

    cs.LG 2026-05 unverdicted novelty 7.0

    A framework combining VQ-VAE, task-relevance scoring, DRL subset selection, and a learned semantic-aware constellation claims near-100% semantic protection probability and proves standard Gray-coded constellations are...

  22. Quantitative Linear Logic for Neuro-Symbolic Learning and Verification

    cs.LO 2026-05 unverdicted novelty 7.0

    QLL is a novel logic for neuro-symbolic learning that uses ML-native operations (sum, log-sum-exp) on logits to embed constraints, satisfying most linear logic properties and showing stronger correlation between empir...

  23. QLAM: A Quantum Long-Attention Memory Approach to Long-Sequence Token Modeling

    cs.LG 2026-05 unverdicted novelty 7.0

    QLAM extends state-space models with quantum superposition in the hidden state for linear-time long-sequence modeling and reports consistent gains over RNN and transformer baselines on sequential image tasks.

  24. FeatCal: Feature Calibration for Post-Merging Models

    cs.LG 2026-05 conditional novelty 7.0

    FeatCal reduces feature drift in merged models via layer-wise closed-form calibration on a small dataset, outperforming prior post-merging methods on CLIP and GLUE benchmarks with high sample efficiency.

  25. From Compression to Accountability: Harmless Copyright Protection for Dataset Distillation

    cs.CR 2026-05 unverdicted novelty 7.0

    SubPopMark protects distilled datasets by injecting verifiable subpopulation biases that create distinguishable model behaviors for copyright tracing without using backdoors.

  26. From Compression to Accountability: Harmless Copyright Protection for Dataset Distillation

    cs.CR 2026-05 unverdicted novelty 7.0

    SubPopMark embeds verifiable subpopulation biases into distilled datasets via CVM and USTM optimization stages, allowing provenance inference through comparison of model output signatures against a reference behavior bank.

  27. Implicit Neural Optimal Transport via Fixed-Point Optimization

    math.OC 2026-05 unverdicted novelty 7.0

    A single-network fixed-point formulation for neural optimal transport eliminates adversarial min-max optimization and implicit differentiation while enforcing dual feasibility exactly.

  28. Implicit Neural Optimal Transport via Fixed-Point Optimization

    math.OC 2026-05 unverdicted novelty 7.0

    A single-network implicit neural optimal transport method that solves the c-transform via proximal fixed-point iteration for stable, non-adversarial training.

  29. Classification-Head Bias in Class-Level Machine Unlearning: Diagnosis, Mitigation, and Evaluation

    cs.LG 2026-05 conditional novelty 7.0

    Class-level unlearning shortcuts via bias suppression in the classification head; new bias-aware training mechanisms and bias-specific metrics are introduced to diagnose and reduce this dependence.

  30. Pre-training Enables Extraordinary All-optical Image Denoising

    physics.optics 2026-05 unverdicted novelty 7.0

    Pre-training diffractive optical networks on millions of simple images followed by fine-tuning enables all-optical denoising that raises PSNR from below 8 dB to above 18 dB across diverse datasets including MNIST, Che...

  31. TRACE: Transport Alignment Conformal Prediction via Diffusion and Flow Matching Models

    stat.ML 2026-05 unverdicted novelty 7.0

    TRACE creates valid conformal prediction sets for complex generative models by scoring outputs via averaged denoising or velocity errors along stochastic transport paths instead of likelihoods.

  32. Test-Time Compositional Generalization in Diffusion Models via Concept Discovery

    cs.LG 2026-05 unverdicted novelty 7.0

    Diffusion models can extract reusable density-mode concepts from their time-indexed scores to enable compositional generation at test time on held-out benchmarks from ColorMNIST and CelebA.

  33. The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models

    stat.ML 2026-05 unverdicted novelty 7.0

    Higher-variance classes are learned first in diffusion models; strong class imbalance reverses the order and imposes distinct delayed learning times on minority classes.

  34. Non-Myopic Active Feature Acquisition via Pathwise Policy Gradients

    cs.LG 2026-05 unverdicted novelty 7.0

    NM-PPG optimizes non-myopic acquisition policies for costly features by enabling pathwise gradients via continuous relaxation and straight-through rollouts in POMDPs, outperforming SOTA baselines.

  35. Spectral Graph Sparsification Preserves Representation Geometry in Graph Neural Networks

    cs.LG 2026-05 unverdicted novelty 7.0

    Spectral sparsification preserves GNN embedding geometry up to O(ε) perturbations in filters, representations, Gram matrices, and training trajectories.

  36. Quantum Interval Bound Propagation for Certified Training of Quantum Neural Networks

    quant-ph 2026-05 unverdicted novelty 7.0

    QIBP adapts interval bound propagation to quantum neural networks for certified adversarial robustness via interval and affine arithmetic implementations.

  37. Heterogeneous-Horizon Exact-Weight Local SGD

    math.OC 2026-04 unverdicted novelty 7.0

    HEW-Local SGD provides exact-weight adaptive aggregation for heterogeneous local SGD with one-step guarantees and explicit convergence results under unequal local horizons.

  38. Diverse Dictionary Learning

    cs.LG 2026-04 unverdicted novelty 7.0

    Diverse dictionary learning identifies intersections, complements, and dependency structures of latent variables from data X = g(Z) up to indeterminacies, and full identifiability when structural diversity is sufficient.

  39. The Multi-Block DC Function Class: Theory, Algorithms, and Applications

    math.OC 2026-04 unverdicted novelty 7.0

    The Multi-Block DC class admits polynomial-size DC decompositions for problems that require exponential size under standard DC programming and supplies explicit constructive formulations for deep ReLU networks togethe...

  40. Feature-level analysis and adversarial transfer in rotationally equivariant quantum machine learning

    quant-ph 2026-04 unverdicted novelty 7.0

    Rotationally equivariant quantum models can rely on vulnerable invariant statistics such as ring-averaged intensities, leaving them susceptible to classical transfer attacks, but suppressing the associated symmetry se...

  41. The Linear Centroids Hypothesis: Features as Directions Learned by Local Experts

    cs.LG 2026-04 unverdicted novelty 7.0

    The Linear Centroids Hypothesis reframes network features as directions in centroid spaces of local affine experts, unifying interpretability methods and yielding sparser, more faithful dictionaries, circuits, and sal...

  42. Tensor-based Multi-layer Decoupling

    eess.SY 2026-04 unverdicted novelty 7.0

    A new tensor framework for multi-layer decoupling of multivariate functions is proposed via ParaTuck decompositions and bilevel optimization.

  43. Toward Exact Convergence in Byzantine-Robust Decentralized Learning: A Statistical Identification Approach

    stat.ME 2026-04 unverdicted novelty 7.0

    DRSGD-ByMI identifies Byzantine machines via sample-splitting score statistics with FDR control, then prunes them to recover sufficient connectivity and achieve order-optimal convergence rates identical to standard de...

  44. XFED: Non-Collusive Model Poisoning Attack Against Byzantine-Robust Federated Classifiers

    cs.CR 2026-04 unverdicted novelty 7.0

    XFED is the first aggregation-agnostic non-collusive model poisoning attack that bypasses eight state-of-the-art defenses on six benchmark datasets without attacker coordination.

  45. Instance-Adaptive Parametrization for Amortized Variational Inference

    cs.LG 2026-04 unverdicted novelty 7.0

    IA-VAE augments amortized variational inference with hypernetwork-generated instance-adaptive modulations, strictly containing the standard variational family and improving held-out ELBO on synthetic and image data.

  46. Drifting Fields are not Conservative

    cs.LG 2026-04 conditional novelty 7.0

    Drift fields in single-pass generative models are not conservative except for Gaussian kernels; a sharp kernel normalization makes them conservative for any radial kernel while noting that non-conservative fields offe...

  47. Selectivity and Shape in the Design of Forward-Forward Goodness Functions

    cs.LG 2026-03 unverdicted novelty 7.0

    Shape- and peak-sensitive goodness functions for Forward-Forward deliver up to 72pp gains over sum-of-squares, reaching 98.2% on MNIST and 89% on Fashion-MNIST.

  48. How Out-of-Equilibrium Phase Transitions can Seed Pattern Formation in Trained Diffusion Models

    cs.LG 2026-03 unverdicted novelty 7.0

    Pattern formation in trained diffusion models emerges from out-of-equilibrium phase transitions driven by instabilities in low-frequency denoising modes linked to data symmetries and architectural constraints.

  49. Programmable superconducting neuron with intrinsic in-memory computation and dual-timescale plasticity for ultra-efficient neuromorphic computing

    cs.ET 2026-03 unverdicted novelty 7.0

    A programmable superconducting LIF neuron with intrinsic static memory and dual-timescale plasticity achieves 45 GHz operation and femtojoule energy per spike.

  50. FlashSinkhorn: IO-Aware Entropic Optimal Transport on GPU

    cs.LG 2026-02 conditional novelty 7.0

    FlashSinkhorn delivers up to 32x forward and 161x end-to-end speedups for entropic OT on A100 GPUs via IO-aware Triton kernels that fuse log-domain updates and streaming transport application.

  51. Re-Key-Free, Risky-Free: Adaptable Model Usage Control

    cs.CR 2025-11 unverdicted novelty 7.0

    AdaLoc keeps a model locked to authorized users by confining all post-deployment updates to a chosen subset of weights, preserving both task performance for authorized use and near-random accuracy for unauthorized use...

  52. Tensor Computation of Euler Characteristic Functions and Transforms

    cs.CG 2025-11 unverdicted novelty 7.0

    A GPU-optimized tensor method computes WECT and ECF for arbitrary-dimensional simplicial and cubical complexes with reported speedups over prior approaches and ships as the pyECT Python package.

  53. RACE Attention: A Strictly Linear-Time Attention Layer for Training on Outrageously Large Contexts

    cs.LG 2025-10 unverdicted novelty 7.0

    RACE Attention is a strictly linear-time attention mechanism that approximates softmax attention outputs using Gaussian projections and soft LSH to enable training on contexts up to 12 million tokens.

  54. The Tsetlin Machine Goes Deep: Logical Learning and Reasoning With Graphs

    cs.LG 2025-07 unverdicted novelty 7.0

    GraphTM uses message passing on graphs to build nested deep clauses, achieving 3.86% higher accuracy than convolutional TM on CIFAR-10 and competitive results on action tracking, recommendations, and genome sequences.

  55. Conformal-DP: A Density-Aware Mechanism for Differential Privacy over Riemannian Manifolds via Conformal Transformation

    cs.CR 2025-04 unverdicted novelty 7.0

    Conformal-DP applies conformal transformations to create a density-aware DP mechanism on Riemannian manifolds, proving ε-DP and deriving a closed-form geodesic error bound dependent only on density ratio and independe...

  56. Privacy Leakage via Output Label Space and Differentially Private Continual Learning

    cs.LG 2024-11 unverdicted novelty 7.0

    Identifies output label space as a privacy side-channel in DP continual learning, formalizes DP for CL, and demonstrates two mitigation methods yielding higher accuracy than prior work.

  57. CRONOS: Enhancing Deep Learning with Scalable GPU Accelerated Convex Neural Networks

    cs.LG 2024-11 unverdicted novelty 7.0

    CRONOS introduces scalable convex optimization for two-layer neural networks reaching ImageNet scale, with CRONOS-AM extending to arbitrary multi-layer architectures while matching tuned deep learning performance.

  58. DaiMoN: A Decentralized Artificial Intelligence Model Network

    cs.LG 2019-07 unverdicted novelty 7.0

    DaiMoN introduces a decentralized ledger-based network for collaborative ML model improvement with label-hidden proof-of-improvement enabled by a novel learnable Distance Embedding for Labels (DEL) function.

  59. k-GANs: Ensemble of Generative Models with Semi-Discrete Optimal Transport

    stat.ML 2019-07 unverdicted novelty 7.0

    k-GANs trains an ensemble of GANs by mapping point masses to Voronoi tiles of the data distribution using semi-discrete optimal transport and iteratively optimizing both generators and point masses, outperforming base...

  60. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    stat.ML 2018-02 unverdicted novelty 7.0

    UMAP is a novel, scalable manifold learning algorithm for dimension reduction that competes with t-SNE while preserving more global structure and having no embedding dimension restrictions.

Reference graph

Works this paper leans on

6 extracted references · 6 canonical work pages · cited by 234 Pith papers

  1. [1]

    Ciregan, U

    D. Ciregan, U. Meier, and J. Schmidhuber. Multi-column deep neural networks for image classification. In Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on, pages 3642--3649. IEEE, 2012

    work page 2012

  2. [2]

    EMNIST: an extension of MNIST to handwritten letters

    G. Cohen, S. Afshar, J. Tapson, and A. van Schaik. Emnist: an extension of mnist to handwritten letters. arXiv preprint arXiv:1702.05373, 2017

    work page Pith review arXiv 2017

  3. [3]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. Imagenet: A large-scale hierarchical image database. In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pages 248--255. IEEE, 2009

    work page 2009

  4. [4]

    Krizhevsky and G

    A. Krizhevsky and G. Hinton. Learning multiple layers of features from tiny images. 2009

    work page 2009

  5. [5]

    LeCun, L

    Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86 0 (11): 0 2278--2324, 1998

    work page 1998

  6. [6]

    L. Wan, M. Zeiler, S. Zhang, Y. L. Cun, and R. Fergus. Regularization of neural networks using dropconnect. In Proceedings of the 30th international conference on machine learning (ICML-13), pages 1058--1066, 2013

    work page 2013