Oshan Mudannayake

Research

Ongoing

  • Reinforcement Learning Hyperparameter Rigor — Paired-Offset Reevaluation

    RL | HPO | Selection Bias | Diagnostics
    Investigating selection-induced optimism in RL hyperparameter search and proposing a paired-offset reevaluation diagnostic. Two manuscripts under review.
    2025 - Present
  • Wildfire Spread Prediction with Spatio-Temporal Deep Learning

    Spatio-Temporal | Deep Learning | Remote Sensing
    Leveraging spatial and temporal data for enhanced wildfire spread prediction. Manuscript under review.
    2024 - Present

Completed

  • On Privacy-Preserved Machine Learning Using Secure Multi-Party Computing: Techniques and Trends

    ML | SMPC | Privacy
    Comprehensive survey on privacy-preserving machine learning that employs secure multi-party computing.
    Apr. 2022 - 2025
    [Paper]
  • Detecting Water In Visual Image Streams Captured From Unmanned Aerial Vehicles

    CV | UNet | Tensorflow | OpenCV | Python
    Our work attempts at detecting water surfaces using Unmanned Aerial Vehicles (UAV) footage.
    Oct. 2022 - 2023
    [Paper]
  • Modeling and Prediction of Municipal Solid Waste Generation in Sri Lanka using Machine Learning Techniques

    ML | Time series | Darts | Python
    We aimed to model and forecast solid waste generation patterns in cities using machine learning techniques.
    May 2021 - Jul. 2022
    [Paper]
  • Machine Learning based Internet Domain Entity Matching and its Applications

    ML | Scikit-learn | Tensorflow | Python
    We tried to predict malicious domain URLs by using machine learning techniques.
    Aug. 2020 - Dec. 2021
  • Realtime Property Evaluation of Large Streaming Graphs

    Graphs | Graph Summarization | Python
    We investigated massive graph stream summarization techniques and proposed an improved graph sketch; kMatrix.
    Jan. 2019 - Jan. 2020
    [Presentation] [Source] [Paper]
  • Machine Learning over Encrypted Data

    ML | Encryption | SMPC | Tensorflow | Python | Google Cloud Platform
    We added a layer over Tensorflow to facilitate machine learning on encrypted data over a distributed network of machines.
    Jul. 2018 - Jan. 2019
    [Poster]