The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]

By Li Deng. 2012

5,999 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Dacheng Tao
3. Ming-Hsuan Yang
4. Trevor Darrell
5. Xiaogang Wang
6. Jiashi Feng
7. Chen Change Loy
8. Andrew Zisserman

In Evaluation & Benchmarks

1. Dhruv Batra
2. Eric P. Xing
3. Yihao Feng
4. A. Hanna
5. Lawrence Carin
6. Stefan Lee
7. Devi Parikh
8. Blake Camp

Citations in the corpus

• Recent advances in convolutional neural networks (2015, neutral): “and imposes a larger loss for point which violates the margin comparing with L 1 -Loss. [19] investigates and compares the performance of softmax with L 2 -SVMs in deep networks. The results on MNIST [62] demonstrate the superiority of L 2 -SVM over softmax. 3.4.2. Softmax Loss Softmax loss is a com”
• TAPAS: Train-less Accuracy Predictor for Architecture Search (2018, background): “Peephole delivers good performance on MNIST and CIFAR-10, however it has not been designed to transfer knowledge from familiar datasets to unseen ones.”
• Deep Learning: Methods and Applications (2014, neutral): “com/exdb/mnist/ and an analysis in [78].”
• Variational Prototype Replays for Continual Learning (2019, methodology): “Permuted MNIST in incremental domain task is a benchmark task protocol in continual learning (Lee et al., 2017; Lopez-Paz et al., 2017; Zenke et al., 2017) (Figure 1).”
• Explaining Classifiers with Causal Concept Effect (CaCE) (2019, background): “, 2019) is a variant of the MNIST dataset (Deng, 2012) where the foreground digit is colored.”
• Counterfactual Visual Explanations (2019, methodology): “We use the same architecture as in the MNIST experiments, resizing the Omniglot images to match.”
• OpenOOD: Benchmarking Generalized Out-of-Distribution Detection (2022, methodology): “Meanwhile, Texture [48], MNIST [42] and SVHN [55] are considered as far-OOD.”
• Computational Science and Engineering (2019, positive): “To further investigate the scalability of LSRN, we consider a more realistic problem, the classification of handwritten digits from the MNIST dataset [ 5 ].”
• High-Frequency Component Helps Explain the Generalization of Convolutional Neural Networks (2019, methodology): “All these experiments are repeated in MNIST [14], FashionMNIST [63], and a subset of ImageNet [13].”
• Lensless computational imaging through deep learning (2017, methodology): “Instead, we took the approach of training a specific “training class,” which we selected phase objects in the form handwritten digits because a database of these is readily available [29] and widely used in the study of various machine learning problems.”
• Online Task-free Continual Learning with Dynamic Sparse Distributed Memory (2022, methodology): “We evaluated our algorithm under an online class-incremental setting on five commonly used image classification datasets: MNIST dataset [13] is a handwritten digit recognition dataset with 60k training samples and 10k testing samples.”
• A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications (2023, background): “dataset called MNIST with its digit images as input (pixel data) is an instance of a classification problem [25].”
• LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning (2023, methodology): “This line of work includes image classification datasets like MNIST [18], CIFAR [34], and ImageNet [17]; segmentation datasets like Core50 [38]; and natural language understanding datasets like GLUE [67] and SuperGLUE [59].”
• The Intrinsic Dimension of Images and Its Impact on Learning (2021, background): “MNIST SVHN CIFAR-10 ImageNet k = 3 7.5 (0.2) 8.5 (0.1) 11.4 (0.2) 15.4 (0.8) k = 4 9.8 (0.3) 11.6 (0.1) 15.9 (0.2) 19.8 (0.9) k = 5 10.9 (0.4) 13.2 (0.1) 18.3 (0.3) 21.6 (1.0) Table 2: Mean and standard error of estimated intrinsic dimensions for practical datasets under the same resolution 32 × 32 ”
• Decentralized Federated Learning: Fundamentals, State of the Art, Frameworks, Trends, and Challenges (2022, methodology): “[52] 2021 A method for D2D network in industrial IoT devices for improving the convergence of communications Cross-device Decentralized Real MobileNet FedAvg (consensus) Asynchronous Random MNIST [172]”