def plot_2d_cmd(metric: str, id_vent: int, size: int, step: int):
# Load dei grafi
original_graph = utility.load_graph()
immunological_graph = utility.load_graph(
gexf_filename="immunological_graph.gexf")
if metric == "precision":
filename = "plot/" + str(id_vent) + "_" + str(size) + "_" + str(
step) + "_precision2d.plt"
if os.path.isfile(filename):
id_vents, original_list, trained_list = visualize.load_plot2D_from_file(
filename)
else:
# Lista vents/nodes del sottografo
_, id_vents = utility.get_node_vent_chessboard(id_vent, size, step)
for i in range(len(id_vents)):
id_vents[i] = str(int(id_vents[i]) + 1)
# Lista precision su grafo originale
original_list = metrics.get_ppv_list(original_graph, id_vents)
# Lista precision su grafo modificato
trained_list = metrics.get_ppv_list(immunological_graph, id_vents)
# Salvataggio dei risultati del trivector PRIMA l'addestramento
visualize.save_plot2D_on_file(
id_vents, original_list, trained_list, "plot/" + str(id_vent) +
"_" + str(size) + "_" + str(step) + "_precision2d.plt")
elif metric == "recall":
filename = "plot/" + str(id_vent) + "_" + str(size) + "_" + str(
step) + "_recall2d.plt"
if os.path.isfile(filename):
id_vents, original_list, trained_list = visualize.load_plot2D_from_file(
filename)
else:
# Lista vents/nodes del sottografo
_, id_vents = utility.get_node_vent_chessboard(id_vent, size, step)
for i in range(len(id_vents)):
id_vents[i] = str(int(id_vents[i]) + 1)
# Lista recall su grafo originale
original_list = metrics.get_tpr_list(original_graph, id_vents)
# Lista recall su grafo trained
trained_list = metrics.get_tpr_list(immunological_graph, id_vents)
# Salvataggio dei risultati su file
visualize.save_plot2D_on_file(
id_vents, original_list, trained_list, "plot/" + str(id_vent) +
"_" + str(size) + "_" + str(step) + "_recall2d.plt")
sns.lineplot(id_vents, original_list, label="original")
sns.lineplot(id_vents, trained_list, label="trained")
plt.show()
def get_node_from_idvent_in_graph(id_vent):
# Ritorna le coordinate del vent nel DEM
coord_vent = vent_in_dem(id_vent)
G = utility.load_graph()
# Date le coordinate del DEM, restituisce il nodo corrispondente
node = get_id_from_coord(G, coord_vent)
return node
def ascii_barrier(id_vent, propagation_method, edges_list):
utm_filename = "ASCII_grids/" + propagation_method + "_" + str(
id_vent) + "_barrier.txt"
# Creazione Matrice da stampare
M = np.zeros((ROWS, COLS), dtype=int)
G = utility.load_graph()
for u, v in edges_list:
coords = G.nodes[u]["coord_regions"].split("|")[0]
coord_u_x, coord_u_y = conversion.cast_coord_attr(coords)
M[coord_u_x][coord_u_y] = 1
coords = G.nodes[v]["coord_regions"].split("|")[0]
coord_v_x, coord_v_y = conversion.cast_coord_attr(coords)
M[coord_v_x][coord_v_y] = 1
# Scrittura del file
with open(utm_filename, 'w') as utmfile:
# Scrittura dell'header del file (northing...)
for i in range(0, len(header)):
utmfile.write(header[i] + "\n")
# Scrittura valori nel DEM
for x in range(0, ROWS):
for y in range(0, COLS):
if M[x][y] == 0:
utmfile.write("0" + " ")
else:
utmfile.write(str(M[x][y]) + " ")
utmfile.write("\n")
def trivector_cmd(id_vent: str,
neighbor_method: str,
radius: int,
threshold: float,
graph: str,
header=False):
propagation = Propagation()
# Setta i parametri
if os.path.isfile("graph_gexf/" + graph):
propagation.set_Graph(utility.load_graph(graph))
else:
print("Graph doesn't exist, default are loaded")
propagation.set_trivector(threshold)
id_vents = []
propagation_method = "trivector"
neighbor = ""
if neighbor_method != None:
if neighbor_method == "moore" or neighbor_method == "neumann":
id_vents = utility.get_neighborhood(id_vent, neighbor_method,
radius)
neighbor += "_" + neighbor_method + str(radius)
else:
print("You must specify a valid neighbor method: moore / neumann.")
return None
else:
id_vents = [id_vent]
# Esecuzione algoritmo
sparse_matrix = propagation.trivector(id_vents)
# Esportazione in ASCII e calcolo metriche
G = propagation.get_Graph()
utility.visualize_and_metrics(id_vents, propagation_method, neighbor,
sparse_matrix, G, header)
def get_trivector_subgraph(tri_vect, real_vect):
G = utility.load_graph()
# lista di nodi invasi dalla simulazione e da trivector.
list_nodes = []
# scorro il vettore della simulazione reale
for i in range(
0, len(real_vect)
): # il nodo è invaso se l'i-esima posizione del vettore è maggiore di 0
if real_vect[i] > 0:
list_nodes.append(str(i))
# espansione simulazione reale al vicinato
tmp = []
for u in list_nodes:
for v in G.successors(u):
if v not in list_nodes and v not in tmp:
tmp.append(v)
list_nodes += tmp
# scorro il vettore di trivector
for i in range(0, len(tri_vect)):
if tri_vect[i] > 0 and str(
i
) not in list_nodes: # il nodo è invaso se l'i-esima posizione del vettore è maggiore di 0
list_nodes.append(str(i))
# creo un grafo copia di G
sub_G = G.copy()
# cancello tutti i nodi del grafo che non sia nella lista
for node in G.nodes:
if node not in list_nodes:
sub_G.remove_node(node)
return sub_G
def norm_to_graph(np_vect):
G = load_graph()
for u in G.nodes():
G.node[u]["current_flow"] = np_vect.item(int(u))
export_graph(G, "norm_from_col_normalization_graph.gexf", False)
return G
def test(id_vent, distance=4):
propagation = Propagation()
propagation.trivector([id_vent])
G = propagation.get_Graph()
# Conversione vent
id_node = conversion.get_node_from_idvent(int(id_vent))
# Estrazione sottografo
for i in range(len(G.nodes)):
if G.nodes[str(i)]['current_flow'] == 0:
G.remove_node(str(i))
# Cambio pesi
for u, v, data in G.edges(data=True):
if data['trasmittance'] == 0:
G.edges[u, v]['trasmittance'] = math.inf
else:
G.edges[u, v]['trasmittance'] = -math.log(data['trasmittance'])
# Lista nodi città
city_nodes = []
for n in G.nodes:
if G.nodes[n]['priority'] > 0:
city_nodes.append(n)
# Shortest paths
shortest_paths = {}
for n in city_nodes:
shortest_paths[n] = nx.shortest_path(G,
id_node,
n,
weight='trasmittance')
cutted_edges = {}
for n in shortest_paths:
path = shortest_paths[n]
for i in range(distance, len(path) - distance):
edge_id = (path[i - 1], path[i])
if edge_id not in cutted_edges:
propagation.set_Graph(utility.load_graph())
propagation.cut_edges([[path[i - 1], path[i]]])
sparse_matrix = propagation.trivector([id_vent])
G = propagation.get_Graph()
risk = metrics.compute([id_vent], 'trivector', sparse_matrix,
G)[-1]
cutted_edges[edge_id] = risk
print(edge_id, risk)
min_risk = min(cutted_edges, key=cutted_edges.get)
print(min_risk)
def __init__(self):
# --- Parametri default ---
# Trivector:
self.tri_threshold = 0.001
self.tri_t_min = 0.015
self.tri_t_max = 0.015
# Eruption:
self.eru_volume = 1000
self.eru_n_days = 7
self.eru_threshold = 0.15
# Montecarlo:
self.prob_epoch = 100
self.prob_second_chance = 0
self.G = utility.load_graph()
def autocut_cmd(id_vent: str, distance: int, neighbor_method: str, radius: int,
dimension: int, mode: str):
propagation = Propagation()
id_vents = []
propagation_method = "trivector"
neighbor = ""
if neighbor_method != None:
if neighbor_method == "moore" or neighbor_method == "neumann":
id_vents = utility.get_neighborhood(id_vent, neighbor_method,
radius)
neighbor += "_" + neighbor_method + str(radius)
else:
print("You must specify a valid neighbor method: moore / neumann.")
return None
else:
id_vents = [id_vent]
# Esecuzione algoritmo prima di tagliare
no_cutted_sparse = propagation.trivector(id_vents)
G = propagation.get_Graph()
# Decommentare se si vuole stampare le metriche del trivector senza taglio
# utility.visualize_and_metrics(id_vents, propagation_method, no_cutted_sparse, G, False)
# selezione degli archi da tagliare
list_edges = ga.get_edges_to_cut(G, id_vents, distance, dimension, mode)
propagation.set_Graph(utility.load_graph())
# taglio archi selezionati
propagation.cut_edges(list_edges)
# esecuzione trivector dopo il taglio degli archi
cutted_sparse = propagation.trivector(id_vents)
G = propagation.get_Graph()
utility.visualize_and_metrics(id_vents, propagation_method, neighbor,
cutted_sparse, G, False)
mc.ascii_barrier(id_vent, propagation_method + neighbor, list_edges)
from transformers import DistilBertTokenizer, AdamW
from sklearn.metrics import accuracy_score, f1_score, classification_report
import model
from tqdm import tqdm
import prepare_data
import config
import utility
import telegram_bot as bot
print("-" * 80)
device = 'cuda' if cuda.is_available() else 'cpu'
print("Available Device: {}".format(device))
print("-" * 60)
#############################################################################################
graph = utility.load_graph(config.generic_path)
print(graph)
#############################################################################################
# Defining the training function on the 80% of the dataset for tuning the distilbert model
def retrain(model, training_loader, loss_fn, optimizer):
tr_loss = 0
n_correct = 0
nb_tr_steps = 0
nb_tr_examples = 0
model.train()
bot.telegram_bot_sendtext("re-training started for : " +
config.generic_path)
for _, data in enumerate(tqdm(training_loader, 0)):
ids = data['ids'].to(device, dtype=torch.long)
def plot_3d_cmd(metric: str, id_vent: int, size: int, step: int):
# Load dei grafi
original_graph = utility.load_graph()
immunological_graph = utility.load_graph(
gexf_filename="immunological_graph.gexf")
if metric == "precision":
filename = "plot/" + str(id_vent) + "_" + str(size) + "_" + str(
step) + "_precision3d.plt"
if os.path.isfile(filename):
x_coords, y_coords, original_list, trained_list = visualize.load_plot3D_from_file(
filename)
else:
# Lista vents/nodes del sottografo
id_nodes, id_vents = utility.get_node_vent_chessboard(
id_vent, size, step)
# Coords
x_coords = []
y_coords = []
for id_node in id_nodes:
x, y = conversion.cast_coord_attr(
original_graph.nodes[id_node]['coord_regions'])
x_coords.append(x)
y_coords.append(y)
# Lista precision su grafo originale
original_list = metrics.get_ppv_list(original_graph, id_vents)
# Lista precision su grafo trained
trained_list = metrics.get_ppv_list(immunological_graph, id_vents)
# Salvataggio dei risultati su file
visualize.save_plot3D_on_file(
x_coords, y_coords, original_list, trained_list,
"plot/" + str(id_vent) + "_" + str(size) + "_" + str(step) +
"_precision3d.plt")
elif metric == "recall":
filename = "plot/" + str(id_vent) + "_" + str(size) + "_" + str(
step) + "_recall3d.plt"
if os.path.isfile(filename):
x_coords, y_coords, original_list, trained_list = visualize.load_plot3D_from_file(
filename)
else:
# Lista vents/nodes del sottografo
id_nodes, id_vents = utility.get_node_vent_chessboard(
id_vent, size, step)
# Coords
x_coords = []
y_coords = []
for id_node in id_nodes:
x, y = conversion.cast_coord_attr(
original_graph.nodes[id_node]['coord_regions'])
x_coords.append(x)
y_coords.append(y)
# Lista recall su grafo originale
original_list = metrics.get_tpr_list(original_graph, id_vents)
# Lista recall su grafo trained
trained_list = metrics.get_tpr_list(immunological_graph, id_vents)
# Salvataggio dei risultati su file
visualize.save_plot3D_on_file(
x_coords, y_coords, original_list, trained_list,
"plot/" + str(id_vent) + "_" + str(size) + "_" + str(step) +
"_recall3d.plt")
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(x_coords,
y_coords,
original_list,
color='red',
marker='X',
s=50)
ax.scatter(x_coords,
y_coords,
trained_list,
color='blue',
marker='D',
s=50)
for i in range(len(trained_list)):
if trained_list[i] > original_list[i]:
color = 'blue'
else:
color = 'red'
ax.plot([x_coords[i], x_coords[i]], [y_coords[i], y_coords[i]],
[original_list[i], trained_list[i]],
color=color)
plt.show()
def make_graphs():
global CN_GRAPH, ER_GRAPH, UPA_GRAPH
CN_GRAPH = utility.load_graph(GRAPH_URL)
ER_GRAPH = er_graph(1239, 0.002)
UPA_GRAPH = upa_graph(1239, 2)
from Genetic_algorithm import Genetic_algorithm
from Propagation import Propagation
import graph_algorithm as ga
import numpy as np
import utility
import networkx as nx
from scipy import sparse
G = utility.load_graph("graphlow.gexf")
for u in G.nodes():
if "Ragalna" in G.nodes[u]['city_names']:
G.nodes[u]['priority'] = .5
nx.write_gexf(G, "ragalna.gexf")
