Code for "Target Trust Is Budget-Dependent"

This repository contains the code for the paper:

Target Trust Is Budget-Dependent: Source-Anchored Shrinkage for Calibration-Free Spiking EEG Driver Drowsiness Detection

Structure

• code/ — main Python scripts for preprocessing, source training, target-budget adaptation, baselines, analysis, and figure generation
• figure_scripts/ — scripts used to generate the main paper figures
• configs/ — frozen run configurations and provenance notes
• requirements.txt — Python dependencies

Main scripts

• spikemlda_revision_original.py — source training and full-target adaptation pipeline for the q2r SpikeMLDA variant
• spikemlda_baselines.py — baseline implementations (EEGNet, ID3RSNet-style, dense CORAL/MMD/DANN)
• spikemlda_conv.py — convolutional encoder and SNN components
• prepare_dd_database.py / download_dd_database.py — MPD-DF data preparation
• analyze_exact_twin.py / analyze_external_validation.py — analysis helpers
• plot_publication_figures.py / make_bspc_framework_figure.py — figure generation
• benchmark_latency_repeated.py / count_forward_macs.py — efficiency diagnostics

Data

Data are not included; they must be downloaded from the public sources cited in the paper:

• SADT-derived bundle (Cao et al., 2019)
• MPD-DF (Li et al., 2026)

Dependencies

pip install -r requirements.txt

Note

This repository contains code only. The manuscript PDF, LaTeX sources, and supplementary materials are stored separately.