Hexiang(Frank) Hu
Ph.D. Student [at] USC [at]
In Deep.
I am passionte with Machine Learning, Computer Vision as well as Natural Language Processing. I aim to combine the great power of vision and language.


  • Graphics: Shading

    Shading refers to depicting depth perception in 3D models or illustrations by varying levels of darkness. In computer graphics, shading refers to the process of altering the color of an object in 3D scene, based on its angle to lights and its distance from lights to create a photorealistic effect.

  • Graphics: Understanding Local Reflectance Model

    Illumination and reflectance over objects makes image looks real, since light-material interaction in real world caused each point of the object have different colors and shades.This post is written to discuss some factors about illumination and reflectance models(global, local, etc.), and explore how computer graphic deals with illumination and reflectance of objects.

  • Druid: A Real-time Analytical Data Store

    Druid is an open source data store designed for real-time exploratory analytics on large data sets. The system combines a column-oriented storage layout, a distributed, shared-nothing architecture, and an advanced indexing structure to allow for the arbitrary exploration of billion-row tables with sub-second latencies.

  • Dynamo: Amazon's Highly Available Key value Store

    Dynamo is a highly available key-value storage system built by Amazon. It sacrifices a consistency under certain failure scenarios, makes extensive use of object versioning and application-assisted conflict resolution in a manner that provides a novel interface for developers to use.

  • Terminal Configuration

    This Post is just used for keeping personal configuration files online

  • Delay Scheduling: A Simple Technique for Achieving Locality and Fairness in Cluster Scheduling

    This is a paper motivated by the scheduling problem raised in traditional FIFO strategy in data-intensive cluster computing system.

    The proposed methodology is designed to get a good tradeoff point between the conflicts of fairness and data locality, which practically improve response time for small jobs by 5x in a multi-user workload and double throughput in an IO-heavy workload.