📅 2025-02-27 — Session: Thesis Supervision and Dimensionality Analysis
🕒 04:00–05:00
🏷️ Labels: Dimensionality, Thesis, Supervision, Autoencoder, PCA, Feature Selection
📂 Project: Teaching
⭐ Priority: MEDIUM
Session Goal
The session aimed to enhance the understanding of dimensionality in socioeconomic datasets, explore feature selection techniques, and develop a structured thesis supervision plan.
Key Activities
- Explored complexities in defining dimensionality in datasets with mixed variable types, focusing on categorical contributions and advanced dimensionality reduction techniques.
- Discussed encoded representations in autoencoders and the need for higher dimensionality in input features.
- Calculated and analyzed the estimated dimensionality for each variable type (binary, categorical, continuous).
- Developed a thesis plan comparing feature selection techniques (Random Forest, PCA, Autoencoders) for income prediction.
- Planned the integration of time and spatial data into a thesis model to enhance predictions.
- Evaluated the need for Fourier features in time series analysis and analyzed frequency spectrum in autoencoder embeddings.
- Outlined a structured supervision plan for thesis development, focusing on data quality, modeling, and iterative feature selection.
Achievements
- Gained insights into dimensionality reduction and feature selection techniques.
- Developed a comprehensive thesis supervision plan with clear objectives and deliverables.
Pending Tasks
- Further empirical testing of feature selection techniques.
- Implementation of the thesis supervision plan with students.