DeepCloud is designed as an open software-defined ecosystem for researchers at different levels with the following salient features and transformative impacts. It is one of the first massively scalable multi-tenant open cloud platform with full-fledged building blocks and comprehensive shared stores (app, model, knowledge, data) for deep learning research and applications.
This project aims at creating deep architectures inspired by cognitive sciences to under visual scenes either in images or videos. The characteristic of the proposed architecture is that it simplifies the inference using biological plausible marginals (object type and spatial location), which can be learned in an unsupervised way directly from data (i.e. without labels).
GraphBTM is a topic model which is an unsupervised algorithm to understand documents. It learns to discover the latent representation of documents and produce meaning clustering of words in the same topic. The goal of GraphBTM is to overcome the limitations of the Latent Dirichlet Allocation (LDA) which suffers from the data sparsity problem in short text and Biterm Topic Model (BTM) which claims an insufficient whole-corpus topic distribution.
Deep learning (DL) has in recent years been widely used in computer vision and natural language processing (NLP) applications due to its superior performance. However, while images and natural languages are known to be rich in structures expressed, for example, by grammar rules, DL has so far not been capable of explicitly representing and enforcing such structures [Huang18]. In this project, we propose an approach to bridging this gap by exploiting tensor product representations (TPR) [Smolensky90a, Smolensky90b], a structured neural-symbolic model developed in cognitive science, aiming to integrate DL with explicit language rules, logical rules, or rules that summarize the human knowledge about the subject.
The goal of this work is to translate streams of data from individual sensors into a shared manifold-space for joint understanding and processing. This work includes investigation of computational topology for manifold learning, data summarization, and intrinsic dimensionality estimation. In practice, for a given application, processing chains are generally developed for a particular sensor or set of sensors.
The goals of this project include developing the ability to use offline, transferred, and real-time learning using various data sources (intermittent state feedback, cloud) to enable SLAM and related image-based estimation methods.
The goal of this project is to develop a machine-learning-enabled video coding strategy for object detection. Most existing video encoders minimizes distortion under a rate constraint. However, for surveillance video, it is desired for a video encoder to maximize detection probability under a rate constraint. To address this, we will design a new video coding strategy that maximizes object detection probability under a rate constraint. We will locate the information important to object detector, develop Rate-Detection-Optimized framework for mode selection, and design optimized bit rate allocation method.
The goals of this project are to design and develop an Intelligent Task Management Platform with task-oriented assistants for commercial and education services, design and develop an open-domain question answering system, and design and develop a deep neural model for question answering with external Knowledge Base.
The goal of this project is to develop new Deep Learning based high dimensional signal models and prediction tools for immersive visual signal coding.
Traditional application-driven (i.e., fault, SLA, or DoS attack detection) “pull” model (SNMP) based network management is expensive and slow for network problem detection, isolation, and root cause analysis. Especially, the recent federation of novel softwareisation and virtualization architectures, as well as Internet of Things (IoT) technologies, require better management over the heterogeneous systems and services. The project adopts “push” based open source forwarding (streaming) network management technologies by using P4 (Programming Protocol-Independent Packet Processors) Inband Network Telemetry (INT).