A deep-learning-based ontology to guide ontological research – Generative models of large datasets are a powerful tool for modelling, training and querying, but they are also a tool for extracting knowledge from the dataset. Many methods for such queries have been developed, from statistical sampling, to model classification, to learning from large natural datasets, to inference from the data and more. In this paper we propose a new and powerful probabilistic model for querying a large dataset via the Generative Adversarial Network. Our approach is trained and trained using a dataset of millions and millions of queries generated by thousands of people. We make use of supervised learning algorithms to extract useful features for querying the dataset rather than just the query. We show that our model can perform well over the network models, using significantly fewer queries. We call our approach Generative Query Answering: Generative Query Answering Machine (GAN-QA) which is a new general purpose non-parametric generative probabilistic model that can serve as a query-driven and query-driven model. We provide experimental results comparing real world queries generated from different methods and experiments validate our model.
We present an application of deep learning techniques for automatic gesture recognition from facial expressions. Based on convolutional neural networks (CNNs), we propose a novel deep learning model for object detection which addresses the problem of using object detection to detect occluded objects. We leverage a Convolutional Neural Network (CNN) to learn the object features in a CNN, and then train the object detector to find the occluded objects. To test our model on real-world applications, we perform this challenge against three different datasets based on the human test set. The test sets are of different kinds and in each case, object recognition was achieved at different difficulty levels compared to the other hand. Furthermore, we demonstrate that our method is capable of detecting objects that are more similar to the human than to similar objects.
Towards Practical Human-Level Decision Trees
Learning Deep Representations of Graphs with Missing Entries
A deep-learning-based ontology to guide ontological research
Feature Extraction in the Presence of Error Models (Extended Version)
DeepKSPD: Learning to detect unusual motion patterns in videosWe present an application of deep learning techniques for automatic gesture recognition from facial expressions. Based on convolutional neural networks (CNNs), we propose a novel deep learning model for object detection which addresses the problem of using object detection to detect occluded objects. We leverage a Convolutional Neural Network (CNN) to learn the object features in a CNN, and then train the object detector to find the occluded objects. To test our model on real-world applications, we perform this challenge against three different datasets based on the human test set. The test sets are of different kinds and in each case, object recognition was achieved at different difficulty levels compared to the other hand. Furthermore, we demonstrate that our method is capable of detecting objects that are more similar to the human than to similar objects.