We propose introspective convolutional networks (ICN) that emphasize the importance of having convolutional neural networks empowered with generative capabilities. We employ a reclassification-by-synthesis algorithm to perform training using a formulation stemmed from the Bayes theory. Our ICN tries to iteratively: (1) synthesize pseudo-negative samples; and (2) enhance itself by improving the classification. The single CNN classifier learned is at the same time generative --- being able to directly synthesize new samples within its own discriminative model. We conduct experiments on benchmark datasets including MNIST, CIFAR-10, and SVHN using state-of-the-art CNN architectures, and observe improved classification results.

The paper proposes a technique to improve the test accuracy of a discriminative model, by synthesizing additional negative input examples during the training process of the model. The negative example generation process has a Bayesian motivation, and is realized by "optimizing" for images (starting from random Gaussian noise) to maximize the probability of a given class label, a la DeepDream or Neural Artistic Style. These generated examples are added to the training set, and training is halted based on performance on a validation set. Experiments demonstrate that this procedure yields (very modest) improvements in test accuracy, and additionally provides some robustness against adversarial examples. The core idea is quite elegant, with an intuitive picture of using the "hard" negatives generated by the network to tighten the decision boundaries around the positive examples.

We propose introspective convolutional networks (ICN) that emphasize the importance of having convolutional neural networks empowered with generative capabilities. We employ a reclassification-by-synthesis algorithm to perform training using a formulation stemmed from the Bayes theory. Our ICN tries to iteratively: (1) synthesize pseudo-negative samples; and (2) enhance itself by improving the classification. The single CNN classifier learned is at the same time generative --- being able to directly synthesize new samples within its own discriminative model. We conduct experiments on benchmark datasets including MNIST, CIFAR-10, and SVHN using state-of-the-art CNN architectures, and observe improved classification results.