Multi-column Deep Neural Networks for Image Classification
arXiv:1202.2745
Abstract
Traditional methods of computer vision and machine learning cannot match human performance on tasks such as the recognition of handwritten digits or traffic signs. Our biologically plausible deep artificial neural network architectures can. Small (often minimal) receptive fields of convolutional winner-take-all neurons yield large network depth, resulting in roughly as many sparsely connected neural layers as found in mammals between retina and visual cortex. Only winner neurons are trained. Several deep neural columns become experts on inputs preprocessed in different ways; their predictions are averaged. Graphics cards allow for fast training. On the very competitive MNIST handwriting benchmark, our method is the first to achieve near-human performance. On a traffic sign recognition benchmark it outperforms humans by a factor of two. We also improve the state-of-the-art on a plethora of common image classification benchmarks.
20 pages, 14 figures, 8 tables
References in corpus (1)
Cited by in corpus (35)
- Adaptive Computation Time for Recurrent Neural Networks
- Grid Long Short-Term Memory
- Deep Learning Techniques for Future Intelligent Cross-Media Retrieval
- Deep Networks with Internal Selective Attention through Feedback Connections
- Deep Learning of Representations: Looking Forward
- Representation Learning: A Review and New Perspectives
- Recognition of Handwritten Digit using Convolutional Neural Network in Python with Tensorflow and Comparison of Performance for Various Hidden Layers
- Adversarial NLI: A New Benchmark for Natural Language Understanding
- Hermes Attack: Steal DNN Models with Lossless Inference Accuracy
- On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models
- Scale-invariant scale-channel networks: Deep networks that generalise to previously unseen scales
- RoIMix: Proposal-Fusion among Multiple Images for Underwater Object Detection
- Machine Vision for Natural Gas Methane Emissions Detection Using an Infrared Camera
- Pixel-aware Deep Function-mixture Network for Spectral Super-Resolution
- Layered SGD: A Decentralized and Synchronous SGD Algorithm for Scalable Deep Neural Network Training
- One Big Net For Everything
- Self-Supervised Representation Learning for Visual Anomaly Detection
- Learning Optimal Data Augmentation Policies via Bayesian Optimization for Image Classification Tasks
- Structure Learning of Deep Networks via DNA Computing Algorithm
- Maximized Lateral Inhibition in Paired Magnetic Domain Wall Racetracks for Neuromorphic Computing
- Interpreting Deep Models through the Lens of Data
- Symmetries and control in generative neural nets
- Scalable domain adaptation of convolutional neural networks
- Biologically-inspired Salience Affected Artificial Neural Network (SANN)
- Unsupervised Learning-based Depth Estimation aided Visual SLAM Approach
- Predicting Effective Diffusivity of Porous Media from Images by Deep Learning
- Multi-level Feature Fusion-based CNN for Local Climate Zone Classification from Sentinel-2 Images: Benchmark Results on the So2Sat LCZ42 Dataset
- A Genetic Algorithm based Kernel-size Selection Approach for a Multi-column Convolutional Neural Network
- PanDA: Panoptic Data Augmentation
- CTNN: Corticothalamic-inspired neural network
- Discriminative Learning of Similarity and Group Equivariant Representations
- Transferred Fusion Learning using Skipped Networks
- A Skip-connected Multi-column Network for Isolated Handwritten Bangla Character and Digit recognition
- SimpleBooks: Long-term dependency book dataset with simplified English vocabulary for word-level language modeling
- A Novel Hybrid CNN-AIS Visual Pattern Recognition Engine