Daily Journal

24.11.2025

Meeting with Prof. Esteban
Prepared the DIPY development environment
- Installation steps
  - Create a venv and source it (or use uv venv and uv pip …)
  - To build the project
    - sudo dnf install python3-devel
    - pip install -r requirements/build.txt
    - pip install –no-build-isolation -e .
  - To build the documentation
    - pip install sphinx
    - pip install sphinx_gallery
    - pip install numpydoc
    - pip install sphinxcontrib-bibtex
    - pip install sphinx-design
    - pip install grg-sphinx-theme
  - To use the correct formatting
    - pip install –no-build-isolation -e .[style]
    - pip install pre-commit
    - pre-commit install
Started to look at the current gqi predict code and pull request
Continued reading the Nipreps book (tried to fix a latex rendering error unsuccessfully)

Finished reading the Nipreps book
Continued to look at the current gqi code and made small changes on the enh/gqi-predict-dev branch
- Changed GeneralizedQSamplingFit odf function to take a sphere parameter (see commit)
- Changed the GeneralizedQSamplingFit predict function to use clamping to remove negative value predictions (see commit)
Created a small marimo notebook to test and understand the prediction function on the enh/gqi-predict-notebooks branch (see commit)

Changed the GeneralizedQSamplingFit predict function to use the model mode when calling the predict_kernel function. Also made the predict function work for multi-voxels (see commit)
Started gqi predict tests with (see commit)
- Single-voxel basic test
- Single-voxel meaningful test
- Multi-voxel basic test
- Multi-voxel meaningful test
Meeting with Prof. Esteban
Did 2 marimo notebooks on the enh/gqi-predict-notebooks branch (see commit)
- A notebook regarding the roundtrip predictions on already seen data (see notebook)
- A notebook regarding cross voxel predictions (see notebook)
Created the 303.1-semester-project github page (see page)

Started working on the cross validation prediction test
Worked on the project documentation (see documentation)
- Added a quarto theme
- Added sources + daily diary pages
- Added mermaid Gantt diagram
- Worked on the report structure (see report)

Added unseen data prediction tests + Cleaned gqi predict functions and existing tests (see commit)
Refactored the gqi predict and prediction_kernel functions to work according to DIPY’s ODF formula:
\(ODF = S \cdot K\) (see commit)
Wrote the The different needed helper functions and The predict function and it’s implementation sections in the report (see commit)

Continued working on the unseen data prediction notebook (see commit and notebook)
Began the DIPY prediction documentation example (see commit)
Revised predict function according to DIPY conventions (not pushed yet)

Added brain mask + interceptor + b0 volumes separation to gqi_unseen_data_notebook (see commit and notebook)
Commented the shapes of the GeneralizedQSamplingFit predict and prediction_kernel functions (see commit)
Optimized the predict multi voxel case a bit (see commit)
Updated the report

Optimized GeneralizedQSamplingFit prediction_kernel function + Added prediction_kernel test which tests: (see commit)
- Correct output shape
- K @ K_plus @ K ≈ K
- K @ K_plus is symmetric
- K_plus @ K is symmetric
Made GeneralizedQSamplingFit predict function faster by fusing kernels therefore avoiding the explicit ODF calculation (see commit)

Meeting with Prof. Esteban
Removed the gqi2 method and b0 volumes from the prediction tests (see commit)
Removed gqi_cross_voxel_notebook + Removed b0 values from train set in gqi_roundtrip_notebook + Cleaned gqi_unseen_data_notebook (see commit)

Cleaned the GeneralizedQSamplingFit predict roundtrip tests (see commit)
Began the Testing Methodology and Obtained Results sections of the report (see report)
Improved marimo notebooks plot titles (see gqi roundtrip notebook, gqi unseen data notebook and commit)

Continued the report, improved the index page and added references (see commit and project website)
Continued working on the presentation (see commit and presentation)

Cleaned project landing page (see commit and langing page)
Fininshed the report and continued the presentation (see commit, report and presentation)
Added submodules section to README.md (see commit)