def Plot(id, function, detail, algorithm, samples):
name = str(id) + ': ' + type(
function).__name__ + " (" + algorithm.__name__ + ")"
print(name, ":\t\t\tComputing...")
fig = plt.figure(name)
fig.suptitle(name)
ax = fig.gca(projection='3d')
fig.canvas.draw_idle()
# Plot the surface.
full = FullFunction(function, detail)
surf = ax.plot_surface(full.X,
full.Y,
full.Z,
cmap=cm.coolwarm,
linewidth=0,
antialiased=True,
alpha=0.5)
plt.pause(0.001)
# Make data.
#search = []
#FuncAnimation(fig, UpdatePlot, fargs=[algorithm, function, ax, search], interval=100, repeat=False, frames=np.linspace(0, samples))
search = [algorithm(function, fig, ax) for x in range(0, samples)]
smin = GraphData(sys.maxsize, sys.maxsize, sys.maxsize)
smax = GraphData(-sys.maxsize, -sys.maxsize, -sys.maxsize)
points = []
for result in search:
points.append(result)
if smin.Z >= result.Z:
smin = result
if smax.Z < result.Z:
smax = result
for point in points:
color = "#000000"
alpha = 0.6
if point.Z == smax.Z:
color = "#0000FF"
alpha = 1.0
if point.Z == smin.Z:
color = "#FF0000"
alpha = 1.0
ax.scatter(point.X, point.Y, point.Z, c=color, alpha=alpha)
print(name, ":\t\t\tMinimum [", smin.X, ", ", smin.Y, "] = ", smin.Z)
print(name, ":\t\t\tMaximum [", smax.X, ", ", smax.Y, "] = ", smax.Z)
# Customize the z axis.
# ax.set_zlim(data.Min, data.Max)
ax.zaxis.set_major_locator(LinearLocator(10))
ax.zaxis.set_major_formatter(FormatStrFormatter('%.02f'))
# Add a color bar which maps values to colors.
fig.colorbar(surf, shrink=0.5, aspect=5)
def FullFunction(function, detail):
step = (function.Range[1] - function.Range[0]) / detail
x = np.arange(function.Range[0], function.Range[1], step)
y = np.arange(function.Range[0], function.Range[1], step)
x, y = np.meshgrid(x, y)
z = function.Value([x, y])
return GraphData(x, y, z)
def fix_mutation(function, y):
r = []
diff = function.Range[1] - function.Range[0]
for x in y.Data:
if function.Range[0] <= x <= function.Range[1]:
r.append(x)
else:
r.append(x % diff)
r.pop()
r.append(function.Value(r))
return GraphData(r)
def plot_pso(function, detail=1000):
# Setup
random.seed()
plt.ion()
args = []
step = (function.Range[1] - function.Range[0]) / 100
for d in range(0, D-1):
args.append(np.random.uniform(function.Range[0], function.Range[1], NP))
population = []
for i in range(0, NP):
p = []
for d in range(0, D-1):
p.append(args[d][i])
p.append(function.Value(p))
m = Member(p)
m.Velocity = GraphData(np.random.uniform(-1, 1, NP))
population.append(m)
fig = plt.figure("PSO")
fig.suptitle("PSO")
ax = fig.gca(projection='3d')
fig.canvas.draw_idle()
# Plot the surface.
full = FullFunction(function, detail)
surf = ax.plot_surface(full.X, full.Y, full.Z, cmap=cm.coolwarm, linewidth=0, antialiased=True, alpha=0.5)
plt.pause(0.001)
points = []
i = 0
best = population[0].Best
while True:
i += 1
print(i, best)
# Draw population
for point in points:
point.remove()
points.clear()
# Move population
for m in population:
best = move_member(m, best, step, function)
points.append(ax.quiver(m.X, m.Y, m.Z, m.Velocity.X, m.Velocity.Y, m.Velocity.Z, length=step, arrow_length_ratio=0.01, alpha=0.3))
px, py, pz = [[p.X for p in population], [p.Y for p in population], [p.Z for p in population]]
points.append(ax.plot(px, py, pz, 'ro', color='black', alpha=0.5)[0])
points.append(ax.plot([best.X], [best.Y], [best.Z], 'ro', color='red')[0])
plt.pause(0.01)
plt.show()
def move_others(function, population, leader, step):
newPopulation = []
newPopulation.append(leader)
leaderIndex = population.index(leader)
for i, member in enumerate(population):
if i == leaderIndex:
continue
pertubatingVector = []
for d in range(0, D - 1):
r = random.random()
pertubatingVector.append(0 if r <= PRT else 1)
pertubatingVector = Vector(pertubatingVector)
newMember = member
t = step
while t < PATH_LENGTH:
# D-1
vector = member + (member - leader) * t * pertubatingVector
data = []
for v in vector.Data:
if v < function.Range[0] or v > function.Range[1]:
data.append(np.random.uniform(function.Range[0], function.Range[1], 1)[0])
else:
data.append(v)
data.append(function.Value(data))
vector = GraphData(data)
if vector.last() <= newMember.last():
newMember = vector
t += step
newPopulation.append(newMember)
return newPopulation
def move_member(m, best, step, function):
# calc new speed
m.Velocity = m.Velocity + C1 * vector_rand() * (m.Best - m) + C2 * vector_rand() * (best - m)
m.Velocity = GraphData(m.Velocity.Data)
# move member
newPosition = m + step * m.Velocity
newPos = []
for i in range(0, D-1):
if function.Range[0] > newPosition.Data[i]:
newPos.append(function.Range[0])
elif function.Range[1] < newPosition.Data[i]:
newPos.append(function.Range[1])
else:
newPos.append(newPosition.Data[i])
# evaluate
newPos.append(function.Value(newPos))
newPos = GraphData(newPos)
if (newPos.last() <= m.Best.last()):
m.Best = newPos
m.set(newPos.Data)
if (m.Best.last() <= best.last()):
best = m.Best
return best
def cross(function, x, y):
# Get random index
r = random.randrange(0, D - 1)
# Generate indexes
rs = [np.random.random() for d in range(0, D - 1)]
# Set values
z = []
for i, ri in enumerate(rs):
if ri <= CR or i == r:
z.append(y.Data[i])
else:
z.append(x.Data[i])
z.append(function.Value(z))
return GraphData(z)
def plot_soma(function, detail=1000):
# Setup
random.seed()
plt.ion()
args = []
for d in range(0, D-1):
args.append(np.random.uniform(function.Range[0], function.Range[1], NP))
population = []
for i in range(0, NP):
p = []
for d in range(0, D-1):
p.append(args[d][i])
p.append(function.Value(p))
population.append(GraphData(p))
fig = plt.figure("SOMA")
fig.suptitle("SOMA")
ax = fig.gca(projection='3d')
fig.canvas.draw_idle()
# Plot the surface.
full = FullFunction(function, detail)
surf = ax.plot_surface(full.X, full.Y, full.Z, cmap=cm.coolwarm, linewidth=0, antialiased=True, alpha=0.5)
plt.pause(0.001)
points = []
i = 0
while True:
i += 1
print(i, len(population))
# Draw population
for point in points:
point.remove()
points.clear()
px, py, pz = [[p.X for p in population], [p.Y for p in population], [p.Z for p in population]]
points.append(ax.plot(px, py, pz, 'ro', color='black')[0])
plt.pause(0.01)
leader = find_leader(population)
leader = move_leader(function, leader)
population = move_others(function, population, leader, STEP)
plt.show()
def plot_diff_evolution(function, detail=1000):
random.seed()
plt.ion()
args = []
for d in range(0, D - 1):
args.append(np.random.uniform(function.Range[0], function.Range[1],
NP))
population = []
for i in range(0, NP):
p = []
for d in range(0, D - 1):
p.append(args[d][i])
p.append(function.Value(p))
population.append(GraphData(p))
fig = plt.figure("DE")
fig.suptitle("DE")
ax = fig.gca(projection='3d')
fig.canvas.draw_idle()
# Plot the surface.
full = FullFunction(function, detail)
surf = ax.plot_surface(full.X,
full.Y,
full.Z,
cmap=cm.coolwarm,
linewidth=0,
antialiased=True,
alpha=0.5)
plt.pause(0.001)
points = []
new_population = []
i = 0
while True:
i += 1
print(i) #, len(population))
plt.pause(0.01)
# Compute
for x in population:
# Get 3 from population
a, b, c = random.sample(population, k=3)
while x == a or x == b or x == c:
[a, b, c] = random.sample(population, k=3)
# Compute temp vector
y = mutate(a, b, c)
y = fix_mutation(function, y)
# Cross selected
y = cross(function, x, y)
# Select new population
if evaluate(x, y):
new_population.append(y)
else:
new_population.append(x)
# Draw population
for point in points:
point.remove()
points.clear()
px, py, pz = [[p.X for p in population], [p.Y for p in population],
[p.Z for p in population]]
points.append(ax.plot(px, py, pz, 'ro', color='black')[0])
population.clear()
population.extend(new_population)
new_population.clear()
plt.show()
def mutate(a, b, c):
return GraphData((a + F * (b - c)).Data)
import matplotlib.pyplot as plt
import numpy as np
import random
from matplotlib import cm
from mpl_toolkits.mplot3d import Axes3D
Axes3D = Axes3D # pycharm auto import
from Models import GraphData
from Utils import FullFunction
F = GraphData(0.5, 0.5, 0.5)
CR = 0.2
NP = 20
D = 3
# best
def mutate(a, b, c):
return GraphData((a + F * (b - c)).Data)
# binomial
def cross(function, x, y):
# Get random index
r = random.randrange(0, D - 1)
# Generate indexes
rs = [np.random.random() for d in range(0, D - 1)]
# Set values
z = []
for i, ri in enumerate(rs):
if ri <= CR or i == r:
def plot_tlbo(function, detail=1000):
# Setup
random.seed()
plt.ion()
# Create population
args = []
step = (function.Range[1] - function.Range[0]) / 100
for d in range(0, D - 1):
args.append(np.random.uniform(function.Range[0], function.Range[1],
NP))
population = []
for i in range(0, NP):
p = []
for d in range(0, D - 1):
p.append(args[d][i])
p.append(function.Value(p))
f = GraphData(p)
population.append(f)
# Set plots
fig = plt.figure("TLBO")
fig.suptitle("TLBO")
ax = fig.gca(projection='3d')
fig.canvas.draw_idle()
# Plot the surface.
full = FullFunction(function, detail)
surf = ax.plot_surface(full.X,
full.Y,
full.Z,
cmap=cm.coolwarm,
linewidth=0,
antialiased=True,
alpha=0.5)
plt.pause(0.001)
points = []
g = 0
teacher = population[0].copy()
while True:
g += 1
print(g, teacher)
# Draw population
for point in points:
point.remove()
points.clear()
px, py, pz = [[p.X for p in population], [p.Y for p in population],
[p.Z for p in population]]
points.append(ax.plot(px, py, pz, 'ro', color='black', alpha=0.5)[0])
points.append(
ax.plot([teacher.X], [teacher.Y], [teacher.Z], 'ro',
color='red')[0])
plt.pause(0.01)
# Find best teacher
x_mean = Vector([0, 0, 0])
for m in population:
if m.last() < teacher.last():
teacher = m.copy()
x_mean += m
x_mean = 1.0 / len(population) * x_mean
Tf = np.round(np.random.uniform(1, 2))
# Teach
newPopulation = []
for m in population:
if m is teacher:
continue
pos = function.FixRange(
(m + vector_rand() * (teacher - Tf * x_mean)).Data)
pos.append(function.Value(pos))
pos = GraphData(pos)
if pos.last() < m.last():
newPopulation.append(pos)
else:
newPopulation.append(m)
# Learn
for m in newPopulation:
l = random.sample(newPopulation, k=1)[0]
while m is l:
l = random.sample(newPopulation, k=1)[0]
pos = function.FixRange(
(m + vector_rand() * ((m - l) if m.last() <= l.last() else
(l - m))).Data)
pos.append(function.Value(pos))
if pos[len(pos) - 1] < m.last():
m.set(pos)
population = newPopulation
plt.show()