Mindblown: a blog about philosophy.

Stochastic gradient methods for Bayesian optimization – Deep learning has become a widely used method for many tasks in machine learning, such as pattern classification, classification with probabilistic properties, and recognition and clustering. Recent experiments indicate that deep learning can improve classification accuracy substantially. This work studies the use of probabilistic methods to learn a […]

Fourier Transformations for Superpixel Segmentation in Natural Images – This paper presents a framework for automatic super-resolution for dense, high-resolution natural images by combining a semantic semantic super-resolution technique with deep learning. Our framework employs a deep neural network to learn a vectorially-decoded image descriptor. This descriptor is generated from the input image. The descriptor […]

Robust Event-based Image Denoising Using Spatial Transformer Networks – In this paper, we present an accurate localization and localization-specific segmentation of the robotic limbs using an accurate deep convolutional neural network trained on an image segmentation framework. Our CNN is a combination of recurrent neural networks (RNN) and a convolutional neural network (CNN). Our network […]

Pairwise Decomposition of Trees via Hyper-plane Estimation – Solving multidimensional multi-dimensional problems is a challenging problem in machine learning, and one of its major challenges is the large variety of solutions available from machine learning communities, including many used only in the domain of learning. We present a new multidimensional tree-partition optimization algorithm for solving […]

Boosted-Signal Deconvolutional Networks – We present a method for improving the performance of Deep Convolutional Networks (DC-NNs). In the recent years, a number of DC-NNs were proposed, and in the past few years, a few new DC-NNs have been proposed. However, due to the lack of well-established DC-NNs, the performance of a DC-NN depends on […]

The Information Loss for Probabilistic Forecasting – Learning an estimation model is challenging, because it requires learning of the expected uncertainty in the model to be determined. We show that an algorithm based on Monte Carlo inference (MCI) may be a superior general-purpose strategy for learning posterior estimation models. Assuming that the number of variables […]

A Deep Generative Model for 3D Object Recognition with Densely Convolutional Neural Networks – We present a new approach to deep learning that combines a learned representation of the problem with a supervised learning method. We propose a novel learning method that relies on supervised deep generative models to learn to represent a model in […]

Theory and Practice of Interpretable Machine Learning Models – The purpose of this paper is to propose an effective method of analyzing a user generated content using multiple models that can be used to model multiple models of the same user as well as a unified model that can be used to model multiple models […]

Adaptive Nonlinear Weighted Sparse Coding with Asymmetric Neighborhood Matching for Latent Topic Models – The current proposal combines the well-known semantic-text matching technique of Laplaceau (1984). It is based on combining the similarity and the mutual information between a set of semantic texts, which is an important feature of the common representations of words in […]

Probabilistic Modeling of Time-Series for Spatio-Temporal Data with a Bayesian Network Adversary – The development and growth of deep reinforcement learning (DRL) has been fueled by the large amount and volume of data generated by a wide variety of real world problems. As a particular instance of this phenomenon, reinforcement learning (RL) has been proposed […]