> ML_LITERATURE // KRIZHEVSKY-2012-IMAGENET-CLASSIFICATION-DEEP-CNN_v1.0
ImageNet Classification with Deep Convolutional Neural Networks
Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton · Advances in Neural Information Processing Systems (NeurIPS) (2012)
seminal-architecture2012foundationalthirdPartyReproduced
Principal Contribution
Demonstrated dramatic error reduction on ImageNet using deep CNNs trained on GPUs with ReLU, Dropout, and data augmentation.
Operational Relevance
Triggered the modern deep learning revolution and GPU hardware acceleration across the entire technology industry.
Assumptions
- Massive labeled datasets (ImageNet) combined with parallel GPU compute unlock deep hierarchical representation capacity
Limitations
- Heavy fully connected classification heads contained 60M parameters, prone to overfitting without 50% dropout
Connected Algorithms, Architectures & Tools
Related Algorithms:
Related Architectures:
Implementing Libraries: