def run_robots(exercise_options: Dict):
start_coordinates = exercise_options['start']
start = Point(start_coordinates['x'], start_coordinates['y'])
field_options = exercise_options['field']
field_height: int = int(field_options['height'])
field_width: int = int(field_options['width'])
sample_field = Field(field_height, field_width)
sample_robot = SampleRobot('sample_robot', sample_field, start, Colors.RED)
sample_robot.run(*exercise_options['run_options'])
user_field = Field(field_height, field_width)
user_robot = UserRobot('user_robot', user_field, start, Colors.BLUE)
user_robot.run(*exercise_options['run_options'])
result: Dict = {
'correct': validate(user_field, sample_field),
'start': {
'x': sample_robot.initial_position.x,
'y': sample_robot.initial_position.y
},
'sample': dump_robot(sample_robot),
'user': dump_robot(user_robot)
}
# Write result to json file
with open('actions.json', 'w+') as actions_file:
actions_file.write(json.dumps(result))
def get_augmented_iodata(i, o):
values = list(set(np.unique(i)) | set(np.unique(o)))
permutations = list(
set(tuple(np.random.permutation(len(values))) for k in range(5)))
for permutation in permutations:
rv = {k: values[v] for k, v in zip(values, permutation)}
inp = np.asarray([[rv[x] for x in line] for line in i])
out = np.asarray([[rv[x] for x in line] for line in o])
yield Field(inp), Field(out)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#while True:
yield Field(field.data)
def run_robots(exercise_options: Dict):
# Inst. fields for both robots
field = Field(exercise_options['field_height'],
exercise_options['field_width'])
# Inst. robots
user_robot = UserRobot('user_robot', field, exercise_options['start'],
Colors.BLUE)
sample_robot = SampleRobot('sample_robot', field,
exercise_options['start'], Colors.RED)
for step_count in range(max_steps):
user_robot.act(exercise_options['run_options'])
sample_robot.act(exercise_options['run_options'])
result: Dict = {
'correct': False, # validate(user_field, sample_field),
'user': {
'name': user_robot.name,
'start': str(user_robot.initial_position),
'actions_size': len(user_robot.actions),
'actions': list(map(str, user_robot.actions))
},
'sample': {
'name': sample_robot.name,
'start': str(sample_robot.initial_position),
'size': len(sample_robot.actions),
'actions': list(map(str, sample_robot.actions))
}
}
# Write result to json file
with open('actions.json', 'w+') as actions_file:
actions_file.write(json_dump(result, indent=2))
def __fields_equal__(user_field: Field, sample_field: Field,
mind_colors: bool) -> bool:
if user_field.height != sample_field.height or user_field.width != sample_field.width:
return False
for row_index in range(user_field.height):
for cell_index in range(user_field.width):
user_cell = user_field.get_by_coordinates(row_index, cell_index)
sample_cell = sample_field.get_by_coordinates(
row_index, cell_index)
if not __compare_cells__(user_cell, sample_cell, mind_colors):
return False
return True
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#repainter = Repaint(field.data)
prediction_data = np.zeros(field.shape)
#print(self.use_zeros)
graph_data, inputs = GraphFeatureExtractor.prepare_graph_features_for_eval(
field, self.use_zeros)
if inputs.shape[0] < 1:
return field
predictions = []
#print(inputs.shape, len(graph_data), len(self.xgb_classifiers))
#print(len(self.xgb_classifiers), inputs.shape)
for i in range(min(inputs.shape[0], self.ncomponents)):
xgb = self.xgb_classifiers[i]
predictions.append(xgb.predict([inputs[i]]))
#result = repainter(preds).tolist()
for comp, pred in zip(graph_data, predictions):
color = pred
for i, j in comp['pos']:
prediction_data[i, j] = color
yield Field(prediction_data)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
resizer = Resizer(self.m1, self.m2)
result = resizer(field.data)
yield Field(result)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
fractal = Fractal(self.m1, self.m2)
result = fractal(field.data)
yield Field(result)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
repainter = Repaint(field.data)
for prediction in self.predictor.predict(field):
result = self.mirror(prediction.data)
result = repainter(result)
yield Field(result)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
(r, c) = get_subpattern(field.data, wildcard=0)
if self.common_patch is not None:
patch = self.common_patch
else:
patch = self.get_patch(field.data, r, c, True)
#print(patch)
if patch is None or np.any(patch == 0):
yield Field(field.data)
return
result = field.data.copy()
coords = np.where(result == 0)
for (x, y) in zip(*coords):
result[x, y] = patch[x % r, y % c]
yield Field(result)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
#yield field.reconstruct(v)
return
if self.input_pattern is not None:
color_convertor = dict(
set(
list(
zip(*np.stack(
[self.input_pattern,
self.crop_func(field.data)], 0).reshape(2, -1)))))
result = np.asarray([[color_convertor.get(x, x) for x in line]
for line in self.pattern])
result = Field(result)
yield result # field.reconstruct(result)
#return
data = self.crop_func(field.data)
h, w = data.shape
h = min(self.pattern.shape[0], h)
w = min(self.pattern.shape[1], w)
ss = self.pattern[:h, :w]
background_colors = np.unique(data[np.where(ss == 0)])
colormap = {
i: [
c for c in np.unique(data[np.where(ss == i)])
if i == 0 or c not in background_colors
]
for i in np.unique(ss)
}
result = np.zeros(self.pattern.shape, dtype=np.uint8)
result[:h, :w] = data[:h, :w]
for key in colormap:
if key == 0:
continue
value = colormap[key]
if len(value) < 1:
continue
coords = np.where(self.pattern == key)
result[coords] = value[0]
yield Field(result) # field.reconstruct(Field(result))
def validate(self, iodata_list, k=3):
scores = []
for iodata in iodata_list:
pred_scores = []
for res in islice(self.predict(iodata.input_field), k):
score = Field.score(res, iodata.output_field)
pred_scores.append(score)
scores.append(max(pred_scores))
# for p in self.predictors[:3]:
# score = p.validate(iodata_list)
# scores.append(score)
if len(scores) == 0:
return 0.0
return np.mean(scores)
def predict(self, field, return_binary=False):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#repainter = Repaint(field.data)
nrows, ncols = field.shape
prediction_data = np.zeros(field.shape, dtype=np.uint8)
#print(self.use_zeros)
graph_data, inputs = GraphFeatureExtractor.prepare_graph_features_for_eval(
field, self.use_zeros)
if inputs.shape[0] < 1:
#return field
preds_binary = []
else:
preds_binary = self.xgb_binary.predict(inputs)
feat = BTFeatureExtractor.make_features_v2(field)
preds = self.xgb.predict(feat).reshape(nrows, ncols)
preds = preds.astype(int) #.tolist()
#result = repainter(preds).tolist()
prediction_data = preds
if len(preds) > 0 and np.sum(preds) > 0:
for comp, cbin in zip(graph_data, preds_binary):
#color = int(new_col) if cbin > 0.5 else comp['color']
#if cbin > 0.5:
# print("new color", new_col, "old_color", comp['color'])
for i, j in comp['pos']:
if cbin > 0.5:
prediction_data[i, j] = preds[i, j]
else:
prediction_data[i, j] = comp['color']
if return_binary:
print(111)
yield preds_binary, graph_data, Field(prediction_data)
else:
yield Field(prediction_data)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#repainter = Repaint(field.data)
nrows, ncols = field.shape
feat = BTFeatureExtractor.make_features_v3(field,
all_square=self.all_square)
preds = self.xgb.predict(feat).reshape(nrows, ncols)
preds = preds.astype(int) #.tolist()
#preds = field.reconstruct(Field(preds))
#preds = repainter(preds).tolist()
preds = Field(preds)
yield preds
def validate(self, iodata_list, k=3):
if isinstance(iodata_list, IOData):
ps = islice(self.predict(iodata_list.input_field), k)
#print(list(ps))
scores = [Field.score(p, iodata_list.output_field) for p in ps]
if len(scores) < 1:
return 0.0
return max(scores)
scores = []
for iodata in iodata_list:
score = self.validate(iodata)
scores.append(score)
if len(scores) < 1:
return 0.0
#print(scores)
return np.mean(scores)
def process_iodata_input(iodata, pattern, crop_func=crop_data):
i = iodata.input_field
o = iodata.output_field
#bg = {c: 0 for c in o.data[np.where(pattern == 0)]}
bg = dict(list(zip(*np.stack([o.data, pattern], 0).reshape(2, -1))))
# current_id = 1
cropped_data = crop_func(i.data)
# #print(cropped_data, id(i.data))
# for line in cropped_data:
# for x in line:
# if x in bg:
# continue
# bg[x] = current_id
# current_id += 1
#return np.asarray([ [ bg[x] for x in line ] for line in cropped_data ])
data = [[bg.get(x, 0) for x in line] for line in cropped_data]
data = Field(data)
return iodata.reconstruct(data).data
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#while True:
if self.best_Dict is None:
return
n, k = field.shape
answer = np.zeros(field.shape, dtype=field.dtype)
for i in range(n):
for j in range(k):
p1, p2 = get_p1_p2(i, j, n, k, self.best_v, self.best_Q1,
self.best_Q2)
color1 = field.data[i, j]
rule = (p1, p2, color1)
answer[i, j] = self.best_Dict.get(rule, color1)
yield Field(answer)
def predict(self, field):
if isinstance(field, IOData):
for v in self.predict(field.input_field):
yield v
return
#repainter = Repaint(field.data)
prediction_data = np.zeros(field.shape)
graph_data, inputs = GraphFeatureExtractor.prepare_graph_features_for_eval(
field)
preds_binary = self.xgb_binary.predict(inputs)
preds_colors = self.xgb.predict(inputs) #.tolist()
#result = repainter(preds).tolist()
for comp, cbin, new_col in zip(graph_data, preds_binary, preds_colors):
color = int(new_col) if cbin > 0.5 else comp['color']
#if cbin > 0.5:
# print("new color", new_col, "old_color", comp['color'])
for i, j in comp['pos']:
prediction_data[i, j] = color
yield Field(prediction_data)
def get_features(iodata_list,
all_square=False,
features_maker=make_features):
feat = []
target = []
for i, iodata in enumerate(iodata_list):
if isinstance(iodata, IOData):
input_field = iodata.input_field.data
output_field = iodata.output_field.data
else:
input_field, output_field = iodata
input_field = input_field.data
output_field = output_field.data
nrows, ncols = input_field.shape
#output_field = output_field.data
target_rows, target_cols = output_field.shape
if output_field.shape == (
1, 1): #and input_field.shape != output_field.shape:
#print(input_field.shape)
#print(input_field)
output_field = increase2shape(output_field, input_field.shape)
# i = np.asarray([[ np.sum(input_field == i) for i in range(10)]])
# o = output_field
# print(i.shape, o.shape)
# feat.extend(i)
# target.extend(o)
# continue
elif (target_rows != nrows) or (target_cols != ncols):
print('Number of input rows:', nrows, 'cols:', ncols)
print('Number of target rows:', target_rows, 'cols:',
target_cols)
not_valid = 1
return None, None, 1
feat.extend(
features_maker(Field(input_field), all_square=all_square))
target.extend(np.array(output_field).reshape(-1, ))
return np.array(feat), np.array(target), 0
