yoyo
  • Home
  • Projects
  • Posts
  • About Me

Hi, my name is Yourui Guo.

I'm a software developer interested in

Machine Learning and Web Development.

Sound Relaxation Exploration using Reinforcement Learning


music-relax

Sounding the Islamic Garden


music-relax

Music Classification According to Cantometrics Features


music-relax


  • UMAP Notes

    (Originally from https://pair-code.github.io/understanding-umap/) - Dimensionality reduction is a powerful tool for machine learning practitioners to visualize and understand large, high dimensional datasets.

  • Database Interview Questions

    Relational database: RDBMS stores data in the form of a collection of tables and relations can be defined between the common fields of these tables. Most modern database management systems like MySQL, Microsoft SQL Server, Oracle, IBM DB2 and Amazon Redshift are based on RDBMS.

  • Javascript ES6 features

    Arrow functions: Arrows are a function shorthand using the => syntax. Defining variables: let -- cannot be redeclared, declared before use, block scope. const -- cannot be redeclared, updated, block scoped. properties of const object can be updated. Spread operator: three dots (…) to obtain the list of parameters. Template literals: To create multiline strings and perform string interpolation.


  • Highly-motivated and creative M.S. graduate in Computer Science, passionate about Machine Learning and Software Development.
  • Adept at tackling problems with ease and thinking independently.
  • Excited to learn new technologies and collaborate with others.
  • Proficient in a range of modern technologies including Python, C/C++ and JavaScript.
  • Seeking opportunities to software engineer, Machine Learning developer and data scientist.





Sound Relaxation Exploration using Reinforcement Learning

The problem was de ned to determine the subjects' preferences for soundscapes that were able to in- crease their relaxation levels. I addressed the problem with two methods: a traditional search method (real time A*) where the state space was formed as a n-dimensional grid world; and a multi-armed bandit algorithm (UCT) built on a hierarchical tree structure. I compared two methods in simulated experiments and UCT outperformed real time A*. Experimental results on the UCT algorithm were obtained with volunteer subjects.

Github link
Sound Relaxation Exploration using Reinforcement Learning

The goal of this project was to design a web-based application that provided users with a location- dependent sonic experience in the Islamic Garden. As they walked through the garden, they would hear a mix of Iranian classical music: the ambient track, overlaid by located sounds, whose intensity varies according to the visitor's distance from the source. The system tracked a user's location, and mixed sounds in real-time according to where the user currently was within the garden space. Users were able to read information about each track, and listen to any track in "solo" mode or mute any track.

Github link
Sound Relaxation Exploration using Reinforcement Learning

The project aimed at classifying folk music according to Cantometrics features developed by Alan Lo- max. Cantometrics described traditional music from around the world using 37 factors such as nasality, accent and melodic shape. The training data included a library of folk music tracks and a metadata sheet that recorded Cantometrics features as well as music information including regions and cultures. The project developed a classi er (convolutional neural network) to classify folk music based on nasality and region where music tracks were turned into spectrograms for input to the classi er.