Personalized-Search-for-Vector-Databases

Paper code

• Code for plot visualisations for the paper
• Figures for the paper

Computed data

• Computed data from the experiments
• exact_merged.csv - 4.2 Validation of the Approach
• large.csv - 4.3 Scalability of the Approach
• approximates_merged.csv - 4.4 Controlling the Refined Set
• cholesky.csv - 5.2 Experiments

Datasets

• Utilitities for downloading datasets
• laions.sh - downloading LAION dataset
• image-text-extractor.py - CLIP extractor, needs to be adjusted for dataset and CLIP has to be downloaded
• Profimedia dataset can be asked from DISA FI MUNI lab: https://disa.fi.muni.cz/research-directions/software/profiset/

Inputs

• Learned matrices from the previous work: https://link.springer.com/chapter/10.1007/978-3-032-06069-3_13

Implementation

• Code for the implementation and experiments
• Utilities for queries/vectors loading
• The computations were done using Metacentrum and the code is adjusted for using this Grid infrastructure: https://www.metacentrum.cz/en/index.html
• Metacentrum - 4.2 Validation of the Approach
• Metacentrum_full_profimedia - 4.3 Scalability of the Approach
• Metacentrum_laion - 4.3 Scalability of the Approach
• Metacentrum_laion_large - 4.3 Scalability of the Approach
• Metacentrum_approximate - 4.4 Controlling the Refined Set
• Metacentrum_Cholesky - 5.2 Experiments
• .ipynb files - jupyters for trying, but the code for experiments are in Metacentrum folders.