Dynamic Capacity Networks
Explore this paper's citation graph
Summary
The Dynamic Capacity Network is introduced, a neural network that can adaptively assign its capacity across different portions of the input data by combining modules of two types: low-capacity sub-networks and high- capacity sub-nets, which indicate that DCNs are able to drastically reduce the number of computations.
- Type
- preprint
- Published
- 2015-11-24
- Cited by
- 109
- References
- 29
- Access
- Open access
- OpenAlex
- https://openalex.org/W2173038751
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:818973
Keywords
MNIST database, Computer science, Convolutional neural network, Selection (genetic algorithm), Focus (optics)
References
- Theano: new features and speed improvements
- Compressing Deep Convolutional Networks using Vector Quantization
- Distilling the Knowledge in a Neural Network
- Speeding up Convolutional Neural Networks with Low Rank Expansions
- Learning where to look
- Going deeper with convolutions
- Towards End-To-End Speech Recognition with Recurrent Neural Networks
- Blocks and Fuel: Frameworks for deep learning
- Approximate Nearest Neighbor Search by Residual Vector Quantization
- Product Quantization for Nearest Neighbor Search
- Learning to combine foveal glimpses with a third-order Boltzmann machine
- Learning Where to Attend with Deep Architectures for Image Tracking
- Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
- Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
- Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
- Theano: Deep Learning on GPUs with Python
- Model compression
- Reading Digits in Natural Images with Unsupervised Feature Learning
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Spatial Transformer Networks
Cited by
- Fully-Adaptive Feature Sharing in Multi-Task Networks with Applications in Person Attribute Classification
- Spatially Adaptive Computation Time for Residual Networks
- Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer
- Enriched Deep Recurrent Visual Attention Model for Multiple Object Recognition
- Learned Contextual Feature Reweighting for Image Geo-Localization
- Person re-identification using visual attention
- Runtime Neural Pruning
- Top-Down Saliency Detection Driven by Visual Classification
- Learning to skip state updates in recurrent neural networks
- A Survey of Model Compression and Acceleration for Deep Neural Networks
- Learning to Super-Resolve Blurry Face and Text Images
- Model Compression and Acceleration for Deep Neural Networks: The Principles, Progress, and Challenges
- Acceleration of neural network model execution on embedded systems
- Targeted Kernel Networks: Faster Convolutions with Attentive Regularization
- EagleEye: Attack-Agnostic Defense against Adversarial Inputs (Technical Report)
- Adaptive Neural Trees
- Dynamic Channel Pruning: Feature Boosting and Suppression
- Runtime Network Routing for Efficient Image Classification
- Stampnet: Unsupervised Multi-Class Object Discovery
- Enhanced Bayesian Compression via Deep Reinforcement Learning