Our code is based on https://github.com/nicopotyka/Uncertainpy, and we added more functions to carry out our experiments.

QBAF files:
fig1.bag
fig3.bag
fig6.bag
fig8.bag
fig9.bag
fig10.bag

Codes: (located at code_RAE/examples/gradual)
file1 - compute RAEs (Definition 4) in section 4.
file2 - approximate RAEs (Algorithm 1) in section 6.
file3 - generate random acyclic QBAFs with specified number of nodes and edges.
file4 - approximate RAE for one edge.
file5 - plot results for Figure 5 in section 6.
(For cyclic QBAFs, run corresponding files: file1_cyclic, file2_cyclic, file3_cyclic, file4_cyclic, file5_cyclic)

Instructions:
1. Run file1 to get RAEs for fig1.bag, fig3.bag, fig8.bag, fig9.bag， fig10.bag.
2. Run file2 to get approximate RAEs for case study 1 with fig6.bag.
3. Run file3 to get a random QBAF with 15 arguments and 25 edges (output to approx.bag), 
then run file4 (random seed always set to 0) to get approximate RAE for a random edge in approx.bag.
4. Alternatively run file3 and file4 to get 4 different QBAFs.
-with 15 arguments and 25/50/75/100 edges (random seeds for file3: 0/2/2/2).
-with 30 arguments and 100/200/300/400 edges (random seeds for file3: 0/0/0/0).
-with 45 arguments and 245/490/735/980 edges (random seeds for file3: 7/3/3/3).
-with 60 arguments and 440/880/1320/1760 edges (random seeds for file3: 0/0/0/0).
5. Run file5 to plot the convergence for these 4 QBAFs with 'node_edge_diff_15.txt', as shown in Figure 4.
-the data in 'node_edge_diff_15.txt' is generated by Instruction 3 & 4.