def project_exp(points, max_projection, expname, average_neighbors, ptype,
exptype):
write_header(expname)
global max_iter
global min_cost
global totalRun
max_iter = 100
for projections in tqdm(range(1, max_projection)):
costs = []
cost_success = []
timeRec = []
iterations = []
successcount = [0, 0]
successtimeandcost = []
successtimeandcost.append([])
successtimeandcost.append([])
diss = mview.DISS(points) #start method to generate dissimilarities
diss.add_projections(
attributes=projections, Q=ptype
) #specify that you want attribute_number attributes generated from physical 3D-2D example, using cylinder projections
Q = diss.Q #recover projection parameters to use along mview.basic (if using fixed projections, otherwise unnecessary)
if exptype == 'fixed':
Q = Q
else:
Q = None
for i in range(0, totalRun):
#DD = {'nodes' : points, 'attributes' : projections}
mv = mview.basic(diss,
Q=Q,
average_neighbors=average_neighbors,
max_iter=max_iter)
costs.append(mv.cost)
timeRec.append(mv.time)
iterations.append(mv.H['iterations'])
if abs(max_iter - mv.H['iterations']) < 2:
successcount[1] = successcount[1] + 1
else:
successcount[0] = successcount[0] + 1
successtimeandcost[0].append(mv.time)
successtimeandcost[1].append(mv.cost)
process_runs(expname, points, average_neighbors, ptype, successcount,
timeRec, costs, iterations, successtimeandcost,
projections)
def points_exp(max_points, expname, average_neighbors, projection):
write_header(expname)
global max_iter
global min_cost
global totalRun
max_iter = 100
step = 100
T = int(max_points / step) + 1
for n in range(1, T):
points = n * step
costs = []
cost_success = []
timeRec = []
iterations = []
successcount = [0, 0]
successtimeandcost = []
successtimeandcost.append([])
successtimeandcost.append([])
D = get_D_3projections(points)
for i in range(0, 10):
mv = mview.basic(D,
Q=projection,
average_neighbors=average_neighbors,
max_iter=max_iter)
costs.append(mv.cost)
timeRec.append(mv.time)
iterations.append(mv.H['iterations'])
if abs(max_iter - mv.H['iterations']) < 2:
successcount[1] = successcount[1] + 1
else:
successcount[0] = successcount[0] + 1
successtimeandcost[0].append(mv.time)
successtimeandcost[1].append(mv.cost)
process_runs(expname, points, average_neighbors, projection,
successcount, timeRec, costs, iterations,
successtimeandcost, len(mv.Q))
def neighbor_exp(points, max_neighbors, expname, projections):
write_header(expname)
global max_iter
global min_cost
global totalRun
for ne in range(0, int(np.log2(max_neighbors))):
average_neighbors = 2**ne
costs = []
cost_success = []
timeRec = []
iterations = []
successcount = [0, 0]
successtimeandcost = []
successtimeandcost.append([])
successtimeandcost.append([])
for i in range(0, 10):
D = get_D_3projections(points)
mv = mview.basic(D,
Q=projections,
average_neighbors=average_neighbors,
max_iter=max_iter,
min_cost=min_cost,
estimate=False)
costs.append(mv.cost)
timeRec.append(mv.time)
iterations.append(mv.H['iterations'])
if abs(max_iter - mv.H['iterations']) < 2:
successcount[1] = successcount[1] + 1
else:
successcount[0] = successcount[0] + 1
successtimeandcost[0].append(mv.time)
successtimeandcost[1].append(mv.cost)
process_runs(expname, points, average_neighbors, projections,
successcount, timeRec, costs, iterations,
successtimeandcost, len(mv.Q))
def get_MPSE(labels):
labels = pd.read_csv("label.csv")['label'].values
p = pd.read_csv("tsne.csv")
p = p[['x', 'y']]
d1 = distance_matrix(p.values, p.values)
p = pd.read_csv("umap.csv")
p = p[['x', 'y']]
d2 = distance_matrix(p.values, p.values)
# p=pd.read_csv("mds.csv").values
# d3=distance_matrix(p,p)
mv = mview.basic([d1, d2],
Q="standard",
average_neighbors=2,
max_iter=500,
verbose=2)
data = np.vstack((mv.X.T, labels.T)).T
df = pd.DataFrame(data=data, columns=("x", "y", "z", "label"))
draw_plot(df[["x", "y", "label"]], 'x', 'y', 'label', "1")
draw_plot(df[["y", "z", "label"]], 'y', 'z', 'label', "2")
draw_plot(df[["z", "x", "label"]], 'z', 'x', 'label', "3")
return mv.X
print(
"Please provide either -d ( distace matices) or -ds (preloaded dataset)"
)
exit(1)
print("<h1>Please keep the window running</h1>")
D = get_matrix(args)
args.projection_type = None if args.projection_type == 'variable' else args.projection_type
# import pdb; pdb.set_trace()
if args.X0 == "True":
args.X0 = True
else:
args.X0 = False
# mv = mview.basic(D,visualization_method = 'tsne', Q=args.projection_type, verbose=2, smart_initialize=args.X0,
# max_iter=args.max_iters, average_neighbors=args.average_neighbors, min_cost=args.min_cost)
mv = mview.basic(D,
verbose=2,
visualization_args={'perplexity': 100},
max_iters=200,
visualization_method='tsne')
# mv = mview.basic(D, verbose=2, max_iters=200)
mv.plot_computations()
# mv.plot_images()
# write_output(mv, args)
parser.add_argument('-pcolor',
'--pcolor',
default='red',
help="vis: poing color",
required=False)
args = parser.parse_args()
if args.preloaded_dataset == None and args.d == None:
print(
"Please provide either -d ( distace matices) or -ds (preloaded dataset)"
)
exit(1)
print("<h1>Please keep the window running</h1>")
D = get_matrix(args)
args.projection_type = None if args.projection_type == 'variable' else args.projection_type
if args.X0 == "True":
args.X0 = True
else:
args.X0 = False
mv = mview.basic(D,
Q=args.projection_type,
verbose=2,
smart_initialize=args.X0,
max_iter=args.max_iters,
average_neighbors=args.average_neighbors,
min_cost=args.min_cost)
write_output(mv, args)