Dreaming to Distill: Data-free Knowledge Transfer via DeepInversion

Abstract

We introduce DeepInversion, a new method for synthesizing images from the image distribution used to train a deep neural network. We 'invert' a trained network (teacher) to synthesize class-conditional input images starting from random noise, without using any additional information about the training dataset. Keeping the teacher fixed, our method optimizes the input while regularizing the distribution of intermediate feature maps using information stored in the batch normalization layers of the teacher. Further, we improve the diversity of synthesized images using Adaptive DeepInversion, which maximizes the Jensen-Shannon divergence between the teacher and student network logits. The resulting synthesized images from networks trained on the CIFAR-10 and ImageNet datasets demonstrate high fidelity and degree of realism, and help enable a new breed of data-free applications - ones that do not require any real images or labeled data. We demonstrate the applicability of our proposed method to three tasks of immense practical importance -- (i) data-free network pruning, (ii) data-free knowledge transfer, and (iii) data-free continual learning. Code is available at https://github.com/NVlabs/DeepInversion

References in corpus (16)

• Distilling the Knowledge in a Neural Network
• Overcoming catastrophic forgetting in neural networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Spectral Normalization for Generative Adversarial Networks
• Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer
• Do Deep Nets Really Need to be Deep?
• Continual Learning Through Synaptic Intelligence
• Mixed Precision Training
• Trained Ternary Quantization
• Visualizing Deep Convolutional Neural Networks Using Natural Pre-Images
• Data-Free Knowledge Distillation for Deep Neural Networks
• Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers
• Few-shot learning of neural networks from scratch by pseudo example optimization
• Dream Distillation: A Data-Independent Model Compression Framework
• Adversarial Neural Network Inversion via Auxiliary Knowledge Alignment
• Hardware-Guided Symbiotic Training for Compact, Accurate, yet Execution-Efficient LSTM

Cited by in corpus (10)

• Privacy-Preserving Blockchain-Based Federated Learning for IoT Devices
• Source Free Domain Adaptation with Image Translation
• RAIN: RegulArization on Input and Network for Black-Box Domain Adaptation
• FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining
• Black-Box Ripper: Copying black-box models using generative evolutionary algorithms
• NaturalInversion: Data-Free Image Synthesis Improving Real-World Consistency
• Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Data
• Adversarial Self-Supervised Data-Free Distillation for Text Classification
• Enhancing Data-Free Adversarial Distillation with Activation Regularization and Virtual Interpolation
• Impression Space from Deep Template Network