Jifeng Dai1, Yang Lu 2, and Ying Nian Wu2
1 Microsoft Research, Asia 2 University of California, Los Angeles (UCLA), USA
This paper investigates generative modeling of the convolutional neural networks (CNNs). The main contributions include: (1) We construct a generative model for CNNs in the form of exponential tilting of a reference distribution. (2) We propose a generative gradient for pre-training CNNs by a non-parametric importance sampling scheme, which is fundamentally different from the commonly used discriminative gradient, and yet has the same computational architecture and cost as the latter. (3) We propose a generative visualization method for the CNNs by sampling from an explicit parametric image distribution. The proposed visualization method can directly draw synthetic samples for any given node in a trained CNN by the Hamiltonian Monte Carlo (HMC) algorithm, without resorting to any extra hold-out images. Experiments on the ImageNet benchmark show that the proposed generative gradient pre-training helps improve the performances of CNNs, and the proposed generative visualization method generates meaningful and varied samples of synthetic images from a large and deep CNN.
Materials can be downloaded from Paper | Tex | Poster.
@inproceedings{Dai2015ICLR, author = {Dai, Jifeng and Lu, Yang and Wu, Ying Nian}, title = {Generative Modeling of Convolutional Neural Networks}, booktitle = {ICLR}, year = {2015} }
The code can be downloaded from here.