Unsupervised Learning with Randomized Labelings – Randomization is generally regarded as a problem of finding an optimal policy that optimizes the information for a given policy. In this paper, we explore how randomized policy optimization can be performed by minimizing the cost function of an unknown policy in terms of the objective function itself, under the assumption that the policy optimizes in the expected (or the unobserved) direction. The expected cost function itself can provide an information-theoretic explanation for this knowledge-theoretic assumption, and thus provides a framework and empirical results for estimating cost functions for unknown policy optimization problems.

Image classification is a challenging problem due to the wide variation of images used in many image processing applications. In each particular problem, researchers have to make use of various techniques such as supervised learning, multilevel learning, and machine learning. The problem is usually characterized by one of two major characteristics: a) image quality is highly variable, and b) it is difficult to estimate the image quality in terms of the true class labels. Therefore, a novel approach is to combine a supervised and a supervised image classification to gain a better and better classification performance. In this paper, we propose and evaluate an unsupervised Deep Reinforcement Learning (DRL) method which combines a supervised and a supervised image classification with a reinforcement learning (RL) method: (1) the RL method learns a model of an image, and (2) the RL method can learn a high-dimensional representation of the image with more accuracy than the supervised model, by training the RL model to classify it. We demonstrate our method on the ILSVRC 2017 and ILSVRC 2012 benchmark datasets.

Adversarial Encoder Encoder

Risk-sensitive Approximation: A Probabilistic Framework with Axiom Theories

Unsupervised Learning with Randomized Labelings

Learning Representations in Data with a Neural Network based Model for Liquor Stores

Towards a unified view on image quality assessmentImage classification is a challenging problem due to the wide variation of images used in many image processing applications. In each particular problem, researchers have to make use of various techniques such as supervised learning, multilevel learning, and machine learning. The problem is usually characterized by one of two major characteristics: a) image quality is highly variable, and b) it is difficult to estimate the image quality in terms of the true class labels. Therefore, a novel approach is to combine a supervised and a supervised image classification to gain a better and better classification performance. In this paper, we propose and evaluate an unsupervised Deep Reinforcement Learning (DRL) method which combines a supervised and a supervised image classification with a reinforcement learning (RL) method: (1) the RL method learns a model of an image, and (2) the RL method can learn a high-dimensional representation of the image with more accuracy than the supervised model, by training the RL model to classify it. We demonstrate our method on the ILSVRC 2017 and ILSVRC 2012 benchmark datasets.