def runWorld(number_of_cells, number_of_food, number_of_preds, canvas, speedSlider, thresholdSlider, shouldDraw, shouldGradient, canvas2):
Cell.useSecondary = number_of_preds > 0
size_of_world = 900
cells = [Cell(0) for i in range(number_of_cells)] # creates the initial cells
preds = [Cell(1) for i in range(number_of_preds)]
food = []
best = []
avg = []
for k in range(number_of_food - len(food)): # creates the needed amount of food
food.append(Food())
j = 0
bestCellNN = None
bestCellFitness = -1
bestPredNN = None
bestPredFitness = -1
while j < 100000000:
j += 1
#if j == 40000:
# number_of_preds = 1
# number_of_cells = 1
# cells = []
# preds = []
#if j >= 40000:
# if len(cells) == 0:
# temp = Cell(0)
# temp.nn = copy.deepcopy(bestCellNN)
# cells.append(temp)
# if len(preds) == 0:
# temp = Cell(1)
# temp.nn = copy.deepcopy(bestPredNN)
# preds.append(temp)
# preds[0].age = 0
# preds[0].energy = 1000
# cells[0].age = 0
# cells[0].energy = 1000
while len(food) < number_of_food:
food.append(Food())
repopulate(cells, number_of_cells, 0)
repopulate(preds, number_of_preds, 1)
totalFitness = 0
for cell in cells:
if cell.fitness > bestCellFitness:
bestCellFitness = cell.fitness
bestCellNN = cell.nn
totalFitness += cell.fitness
for cell in preds:
if cell.fitness > bestPredFitness:
bestPredFitness = cell.fitness
bestPredNN = cell.nn
spec = Cell(0)
replaced = False
for z in cells:
if z.alive:
if(z.fitness > spec.fitness):
spec = z
replaced = True
if j % 1000 == 0:
print(str(j) + " " + str(spec.fitness) + " " + str(totalFitness))
best.append(spec.fitness)
avg.append(totalFitness/number_of_cells)
plt.plot(best,'g')
plt.plot(avg,'r')
plt.savefig('stats.svg')
canvas.update()
if shouldDraw.get() == 1:
canvas.delete("all")
canvas2.delete("all")
if shouldGradient.get():
bestCopy = Cell(0)
bestCopy.nn = copy.deepcopy(spec.nn)
for x in range(0, 900, 50):
for y in range(0, 900, 50):
bestCopy.direction = 0
bestCopy.x_pos = x
bestCopy.y_pos = y
for i in range(1,10):
tmpX = bestCopy.x_pos
tmpY = bestCopy.y_pos
bestCopy.update_closests(food, preds, True)
bestCopy.make_move(food, preds)
canvas.create_line(tmpX,tmpY,bestCopy.x_pos,bestCopy.y_pos)
for f in food:
canvas.create_rectangle(f.x_pos-3, f.y_pos-3, f.x_pos+3, f.y_pos+3, fill = "black")
for c in cells:
if c.fitness >= thresholdSlider.get():
canvas.create_circle(c.x_pos, c.y_pos, 6, fill = '#%02x%02x%02x' % (c.color[0], c.color[1], c.color[2]))
canvas.create_text(c.x_pos-7,c.y_pos-13, text='%de / %df / %.1fs' % (c.energy, c.fitness, c.speed))
for c in preds:
canvas.create_circle(c.x_pos, c.y_pos, 6, fill = 'black')
if not replaced:
replaced = True
spec = cells[0]
if replaced:
canvas.create_circle(spec.x_pos, spec.y_pos, 2, fill = "red")
fieldNames = ['weights_input_hidden1','weights_hidden1_hidden2', 'weights_hidden2_output']
inputs = spec.get_inputs(food, preds)
inputs = numpy.transpose(numpy.array(inputs, ndmin=2).T)
ptr = 0
gapY = 200
gapX = 75
rad = 25
for field in fieldNames:
arr = getattr(spec.nn, field)
for a in range(len(arr)):
for b in range(len(arr[a])):
canvas2.create_line(50+gapX*a+(7-len(arr))*gapX/2, 250+gapY*(ptr-1),50+gapX*b+(7-len(arr[a]))*gapX/2, 250+gapY*ptr, width = Cell.logistic(8, 1, 0, arr[a][b]))
ptr += 1
ptr = 0
for i in range(len(inputs[0])):
col = "red"
if(inputs[0][i] < 0):
col = "blue"
value = abs(inputs[0][i] / 1000 * rad)
if numpy.isinf(value) or numpy.isnan(value):
value = rad
else:
value = int(value)
canvas2.create_circle(50+gapX*i+(7-len(inputs[0]))*gapX/2, 250+gapY*-1, value, fill=col)
for field in fieldNames:
arr = getattr(spec.nn, field)
inputs = numpy.dot(inputs, arr)
if not field == 'weights_hidden2_output':
inputs = scipy.special.expit(inputs)
else:
inputs[0][0] %= math.pi*2
inputs[0][0] -= math.pi
inputs[0][0] /= math.pi
if len(inputs[0]) >= 2:
inputs[0][1] = Cell.logistic(1, .3, 0, inputs[0][1])
for i in range(len(inputs[0])):
col = "red"
if(inputs[0][i] < 0):
col = "blue"
value = abs(inputs[0][i]*rad)
if numpy.isinf(value) or numpy.isnan(value):
value = rad
else:
value = int(value)
canvas2.create_circle(50+gapX*i+(7-len(inputs[0]))*gapX/2, 250+gapY*ptr, value, fill=col)
ptr += 1
canvas2.update()
canvas.update()
if multi:
pool = ThreadPool()
pool.map(Runner(food,preds,j),cells)
pool.close()
pool.join()
if not multi:
for z in cells:
if z.alive:
z.single_cycle(food, preds, j)
for z in preds:
if z.alive:
z.single_cycle(cells, food, j)
