Fixed model so that it is reproducible between runs. Total time used: 5h 45m (14.4%)
Spent some time revises how number generation works in Python - notes here. I felt this was external to reproducing the model, so did not include it in time.
If I would have timed it though, it would’ve been apx. 9.30-11.28.
I returned to trying to modify the code so it can reproduce results between runs.
A simple fix of adding a seed to some of the Uniform() functions that were still using random_seed=None meant that the simulation would now run exactly the same each time.
Run 1:
Run 2:
Then we needed a seperate seed with each run (else every run will produce identical results) This required changes to:
single_run() within sim_replicate.py so it can accept random number set as an argument, and then when it was called in multiple_replications(), to use the rep number as the starting seedThese changes were based on this model.
As can see, this has fixed it, as we have reproducible results between runs, and varying results within runs.
Run 1:
Run 2:
Exploring methods for overlaying figures. Not timed as not about reproduction of this study, but about how we are going to do this each time when reproducing.
Decided that it’s not helpful to do this - spend more time fiddling around with getting them to resize and overlay correctly - and that the simplest option here would be to compare by eye.
Base image sourced from Allen et al. (2020)
If timed, 13.33-13.57.
import cv2
import matplotlib.pyplot as plt
original = cv2.imread('../../../original_study/article_fig4.png')
overlay = cv2.imread('figure_2a.png')
# Resize overlay to match original image dimensions
overlay_resize = cv2.resize(overlay, (original.shape[1], original.shape[0]))
# View new image
img = plt.imshow(overlay_resize)
plt.axis('off')
img
# Overlay transparently
combined = cv2.addWeighted(original,0.4,overlay_resize,0.1,0)
# View new image
img = plt.imshow(combined)
plt.axis('off')
img
import sys
sys.path.append('../')
from timings import calculate_times
# Minutes used prior to today
used_to_date = 276
# Times from today
times = [
('11.29', '12.38')]
# Print time used and remaining
calculate_times(used_to_date, times)Time spent today: 69m, or 1h 9m
Total used to date: 345m, or 5h 45m
Time remaining: 2055m, or 34h 15m
Used 14.4% of 40 hours max✅ = Made the change.
💬 = Noted to discuss with team.
Protocol: