ConGen-Results

This repository contains some example datasets generated using ConGen. Furthermore, this repository contains the Apptainer image for ConGen.

Gradient datasets

The datasets in the Gradients-folder are generated by modelling population occurency as a linear gradient around randomly generated center points.

Non-binary datasets

Per feature:

1. draw x-y-coordinate vectors with values in [0, res] from a uniform distribution for population centers in 2D square plane
2. res is the specified resolution of the dataset
3. All pus closer than 10 to a population center get a feature value of 1 assigned
4. All pus within 10 and 30 to the population center get their euklidean distance to the population center as feature value assigned
5. Gradients are cut-off at the borders of the 2D square [0, res]
6. Graph nodes for a full graph are placed on pus with values within [209, 210]
7. Graph edges are weighted by their euclidean length
8. A perlin noise function with 2 octaves is sampled for the quality data of each pu

Binary datasets

Identical as non-binary, but step 4 is skipped. Graph nodes are population centers drawn in step 1 instead

Random walk datasets

The datasets in the Random Walk folder are created by generating population occurency data as random walks starting in randomly generated points.

Description of the data generation process

Per Feature:

• Generate a Simplex Noise layer with values between -500 and 500
• Place 25 random points on planning units in the positive areas of the Simplex noise layer
• Generate occurrence data for this feature by simulating a random walk in each point with a fixed number of steps.
  • For each step, randomly decide whether to go to the neighboring planning unit to the top, bottom, left or right of the current planning unit.
  • For each position, calculate the probabilities for each direction to be proportional to the value in the Simplex Noise layer of the planning unit in that direction, scaled to the interval [0, 1] as follows:
    • If the value of all neighboring planning units of a position is 0, set the probability for each direction to 0.25
    • If any neighboring value is below zero, add the lowest value scaled by an exploration factor (1 in the current dataset) to all values to ensure all values are >= 0
    • Set the probabilities for directions exceeding the borders of the dataset to 0
    • Divide all probabilities of a location by their sum to scale to [0, 1]
  • Each time a planning unit is visited, the value for the current feature in that planning unit is increased by 1.
  • Finally, scale all values for the entire feature to [0, 500] proportionally
• Create a complete graph out of all planning units that have at least a feature value of 200 (40% of maximum)
• Generate a Perlin Noise layer with values in [0, 500] as quality data
• Discard quality values for all planning units not included in the complete graph

400x400 pixels

There's three datasets with a resolution of 400x400 pixels, with 5, 10 and 15 features, 25 random walks with 10.000 steps each. Each feature has a different random seed, starting at 0, increasing by 1. The 15 feature dataset includes all features from the 10 feature dataset, and the 10 feature dataset in turn includes all features from the 5 feature dataset. Therefore, the 15 feature dataset includes preview images for all feature layers. Additionally, each dataset includes a preview image with all feature layers and starting points visible.

Overview

Visualisation of the 400x400 dataset with 15 features

All features

Visualisation of feature 00 in the 400x400 dataset with seed 00 Visualisation of feature 01 in the 400x400 dataset with seed 01 Visualisation of feature 02 in the 400x400 dataset with seed 02 Visualisation of feature 03 in the 400x400 dataset with seed 03 Visualisation of feature 04 in the 400x400 dataset with seed 04 Visualisation of feature 05 in the 400x400 dataset with seed 05 Visualisation of feature 06 in the 400x400 dataset with seed 06 Visualisation of feature 07 in the 400x400 dataset with seed 07 Visualisation of feature 08 in the 400x400 dataset with seed 08 Visualisation of feature 09 in the 400x400 dataset with seed 09 Visualisation of feature 10 in the 400x400 dataset with seed 10 Visualisation of feature 11 in the 400x400 dataset with seed 11 Visualisation of feature 12 in the 400x400 dataset with seed 12 Visualisation of feature 13 in the 400x400 dataset with seed 13 Visualisation of feature 14 in the 400x400 dataset with seed 14

800x800 pixels

Same method as for the 400x400 datasets, but now the number of random walks also increases by 5 per feature to model species with different density, and to cope for the larger area.

Overview

Visualisation of the 800x800 dataset with 5 features Visualisation of the 800x800 dataset with 5 features Visualisation of the 800x800 dataset with 5 features

All features

Visualisation of feature 00 in the 800x800 dataset with seed 00 Visualisation of feature 01 in the 800x800 dataset with seed 01 Visualisation of feature 02 in the 800x800 dataset with seed 02 Visualisation of feature 03 in the 800x800 dataset with seed 03 Visualisation of feature 04 in the 800x800 dataset with seed 04 Visualisation of feature 05 in the 800x800 dataset with seed 05 Visualisation of feature 06 in the 800x800 dataset with seed 06 Visualisation of feature 07 in the 800x800 dataset with seed 07 Visualisation of feature 08 in the 800x800 dataset with seed 08 Visualisation of feature 09 in the 800x800 dataset with seed 09 Visualisation of feature 10 in the 800x800 dataset with seed 10 Visualisation of feature 11 in the 800x800 dataset with seed 11 Visualisation of feature 12 in the 800x800 dataset with seed 12 Visualisation of feature 13 in the 800x800 dataset with seed 13 Visualisation of feature 14 in the 800x800 dataset with seed 14

1200x1200 pixels

Same method as for the 800x800 datasets, but now each random walk has 25.000 instead of 10.000 steps to additionally cope for the larger area.

Overview

Visualisation of the 1200x1200 dataset with 5 features Visualisation of the 1200x1200 dataset with 5 features Visualisation of the 1200x1200 dataset with 5 features

All features

Visualisation of feature 00 in the 1200x1200 dataset with seed 00 Visualisation of feature 01 in the 1200x1200 dataset with seed 01 Visualisation of feature 02 in the 1200x1200 dataset with seed 02 Visualisation of feature 03 in the 1200x1200 dataset with seed 03 Visualisation of feature 04 in the 1200x1200 dataset with seed 04 Visualisation of feature 05 in the 1200x1200 dataset with seed 05 Visualisation of feature 06 in the 1200x1200 dataset with seed 06 Visualisation of feature 07 in the 1200x1200 dataset with seed 07 Visualisation of feature 08 in the 1200x1200 dataset with seed 08 Visualisation of feature 09 in the 1200x1200 dataset with seed 09 Visualisation of feature 10 in the 1200x1200 dataset with seed 10 Visualisation of feature 11 in the 1200x1200 dataset with seed 11 Visualisation of feature 12 in the 1200x1200 dataset with seed 12 Visualisation of feature 13 in the 1200x1200 dataset with seed 13 Visualisation of feature 14 in the 1200x1200 dataset with seed 14