Multisource Machine Learning - CS6804
Fall 2022
Overview
This course will provide an introduction to recent research in the area of multi-source learning. The course will survey recent approaches to model multi-source datasets, focusing on fundamental challenges in representation, learning, and inference. The class will be highly interactive, collaborative, and hopefully fun! The bulk of the work will be a class project and reading and presenting papers.
Learning objectives
- Learn about Multi-source Learning
- Work on a research project in the area of Multi-source Learning
- Sharpen necessary skills for conducting research including: (1) authoring, reviewing, and presenting research papers, (2) defining and authoring small research proposals, (3) reviewing a research topic and authoring a research survey paper, (4) participating in research discussions, and (5) collaborating on research.
Background
The suggested background includes machine learning and data analytics.
Projects
Students are encouraged to propose a project that leverages multi-source learning. Potential projects include: investigating the performance of multi-source learning algorithms, analyzing data with multi-source learning models, designing and implementing multi-source learning model/algorithm extensions, and a survey paper on a chosen problem or class of techniques.
Class Projects
- Road Event Classification with Gaze Information
- Effectiveness of Deep Neural Network for Full Waveform Inversion
- Ensemble Learning-based Prediction Engine for Scheduling Heterogeneous GPUs
- IS CRIME A SOCIAL OR INDIVIDUAL PROBLEM? FACT CHECK BASED ON THE US STATES DATA ANALYSIS
- DEEP MULTIMODAL FUSION NETWORKS FOR IMAGE-TEXT CLASSIFICATION
- Deep Fusion Architectures for Cloud Masking
- Improving Code Translation Quality by Deep Reinforcement Learning using Compiler Feedbacks
- Efficacy of Meta-Learning for Few-Shot Learning on Fine-Grained Visual Recognition
- Forecasting ARG composition in Effluents of Waste Water Treatment Plant
- Robustness Analysis of Multi-modal Fusion against Adversarial Attacks
- Video Misinformation Detection