def __call__(self, state, scope, pos, paramTypes, x, *args):
if len(args) == 3:
numbins, low, high = args
if low >= high or math.isnan(low) or math.isnan(high):
raise PFARuntimeException("bad histogram range", self.errcodeBase + 0, self.name, pos)
if numbins < 1:
raise PFARuntimeException("bad histogram scale", self.errcodeBase + 1, self.name, pos)
if math.isnan(x) or x < low or x >= high:
raise PFARuntimeException("x out of range", self.errcodeBase + 2, self.name, pos)
out = int(math.floor(numbins * div((x - low), (high - low))))
if out < 0 or out >= numbins:
raise PFARuntimeException("x out of range", self.errcodeBase + 2, self.name, pos)
return out
else:
origin, width = args
if math.isnan(origin) or math.isinf(origin):
raise PFARuntimeException("bad histogram range", self.errcodeBase + 0, self.name, pos)
if width <= 0.0 or math.isnan(width):
raise PFARuntimeException("bad histogram scale", self.errcodeBase + 1, self.name, pos)
if math.isnan(x) or math.isinf(x):
raise PFARuntimeException("x out of range", self.errcodeBase + 2, self.name, pos)
else:
return int(math.floor(div((x - origin), width)))
def __call__(self, state, scope, pos, paramTypes, x, alpha, beta, gamma, theState):
if alpha < 0.0 or alpha > 1.0:
raise PFARuntimeException("alpha out of range", self.errcodeBase + 0, self.name, pos)
if beta < 0.0 or beta > 1.0:
raise PFARuntimeException("beta out of range", self.errcodeBase + 1, self.name, pos)
if gamma < 0.0 or gamma > 1.0:
raise PFARuntimeException("gamma out of range", self.errcodeBase + 2, self.name, pos)
level_prev = theState["level"]
trend_prev = theState["trend"]
cycle_unrotated = theState["cycle"]
if len(cycle_unrotated) == 0:
raise PFARuntimeException("empty cycle", self.errcodeBase + 3, self.name, pos)
cycle_rotated = cycle_unrotated[1:] + [cycle_unrotated[0]]
cycle_prev = cycle_rotated[0]
if theState["multiplicative"]:
level = div(alpha * x, cycle_prev) + (1.0 - alpha) * (level_prev + trend_prev)
trend = beta * (level - level_prev) + (1.0 - beta) * trend_prev
cycle = div(gamma * x, level) + (1.0 - gamma) * cycle_prev
else:
level = alpha * (x - cycle_prev) + (1.0 - alpha) * (level_prev + trend_prev)
trend = beta * (level - level_prev) + (1.0 - beta) * trend_prev
cycle = gamma * (x - level) + (1.0 - gamma) * cycle_prev
return dict(theState, level=level, trend=trend, cycle=([cycle] + cycle_rotated[1:]))
def __call__(self, state, scope, paramTypes, x, alpha, beta, gamma, theState):
if alpha < 0.0 or alpha > 1.0:
raise PFARuntimeException("alpha is out of range")
if beta < 0.0 or beta > 1.0:
raise PFARuntimeException("beta is out of range")
if gamma < 0.0 or gamma > 1.0:
raise PFARuntimeException("gamma is out of range")
level_prev = theState["level"]
trend_prev = theState["trend"]
cycle_unrotated = theState["cycle"]
if len(cycle_unrotated) == 0:
raise PFARuntimeException("empty cycle")
cycle_rotated = cycle_unrotated[1:] + [cycle_unrotated[0]]
cycle_prev = cycle_rotated[0]
if theState["multiplicative"]:
level = div(alpha * x, cycle_prev) + (1.0 - alpha) * (level_prev + trend_prev)
trend = beta * (level - level_prev) + (1.0 - beta) * trend_prev
cycle = div(gamma * x, level) + (1.0 - gamma) * cycle_prev
else:
level = alpha * (x - cycle_prev) + (1.0 - alpha) * (level_prev + trend_prev)
trend = beta * (level - level_prev) + (1.0 - beta) * trend_prev
cycle = gamma * (x - level) + (1.0 - gamma) * cycle_prev
return dict(theState, level=level, trend=trend, cycle=([cycle] + cycle_rotated[1:]))
def __call__(self, state, scope, pos, paramTypes, observation, prediction,
uncertainty):
if isinstance(observation, dict):
if set(observation.keys()) != set(prediction.keys()):
raise PFARuntimeException("misaligned prediction",
self.errcodeBase + 0, self.name, pos)
if set(observation.keys()) != set(uncertainty.keys()):
raise PFARuntimeException("misaligned uncertainty",
self.errcodeBase + 1, self.name, pos)
return dict(
(k, div(observation[k] - prediction[k], uncertainty[k]))
for k in observation)
elif isinstance(observation, (tuple, list)):
if len(observation) != len(prediction):
raise PFARuntimeException("misaligned prediction",
self.errcodeBase + 0, self.name, pos)
if len(observation) != len(uncertainty):
raise PFARuntimeException("misaligned uncertainty",
self.errcodeBase + 1, self.name, pos)
return [
float(div(o - p, u))
for o, p, u in zip(observation, prediction, uncertainty)
]
else:
return div(observation - prediction, uncertainty)
def interpolateSingle(datum, one, two):
onex = one["x"]
twox = two["x"]
oney = one["to"]
twoy = two["to"]
unitless = div((datum - onex), (twox - onex))
return (1.0 - unitless)*oney + unitless*twoy
def __call__(self, state, scope, pos, paramTypes, x, y):
length = len(x)
if len(y) != length:
raise PFARuntimeException("dimensions of vectors do not match",
self.errcodeBase + 0, self.name, pos)
a00, a01, a10, a11 = self.countPairs(x, y)
return div((a11 + a00), (a11 + 2 * (a10 + a01) + a00))
def __call__(self, state, scope, pos, paramTypes, x, w, theState, level):
originalCount = theState["count"]
count = originalCount + w
if level == 0:
return dict(theState, count=count)
else:
mean = theState["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
if level == 1:
return dict(theState, count=count, mean=mean)
else:
varianceTimesCount = theState["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
return dict(theState, count=count, mean=mean, variance=div(varianceTimesCount, count))
def __call__(self, state, scope, paramTypes, x, w, theState, level):
originalCount = theState["count"]
count = originalCount + w
if level == 0:
return dict(theState, count=count)
else:
mean = theState["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
if level == 1:
return dict(theState, count=count, mean=mean)
else:
varianceTimesCount = theState["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
return dict(theState, count=count, mean=mean, variance=div(varianceTimesCount, count))
def __call__(self, state, scope, pos, paramTypes, data, cluster, weight):
if len(data) == 0:
raise PFARuntimeException("no data", self.errcodeBase + 0,
self.name, pos)
dimension = len(data[0])
summ = [0.0 for i in xrange(dimension)]
for vec in data:
if len(vec) != dimension:
raise PFARuntimeException("dimensions of vectors do not match",
self.errcodeBase + 1, self.name, pos)
for i in xrange(dimension):
summ[i] += vec[i]
vec = cluster["center"]
if len(vec) != dimension:
raise PFARuntimeException("dimensions of vectors do not match",
self.errcodeBase + 1, self.name, pos)
for i in xrange(dimension):
summ[i] += weight * vec[i]
denom = len(data) + weight
for i in xrange(dimension):
summ[i] = div(summ[i], denom)
return dict(cluster, center=summ)
def __call__(self, state, scope, pos, paramTypes, observation, prediction, uncertainty):
if isinstance(observation, dict):
if set(observation.keys()) != set(prediction.keys()):
raise PFARuntimeException("misaligned prediction", self.errcodeBase + 0, self.name, pos)
if set(observation.keys()) != set(uncertainty.keys()):
raise PFARuntimeException("misaligned uncertainty", self.errcodeBase + 1, self.name, pos)
return dict((k, div(observation[k] - prediction[k], uncertainty[k])) for k in observation)
elif isinstance(observation, (tuple, list)):
if len(observation) != len(prediction):
raise PFARuntimeException("misaligned prediction", self.errcodeBase + 0, self.name, pos)
if len(observation) != len(uncertainty):
raise PFARuntimeException("misaligned uncertainty", self.errcodeBase + 1, self.name, pos)
return [float(div(o - p, u)) for o, p, u in zip(observation, prediction, uncertainty)]
else:
return div(observation - prediction, uncertainty)
def __call__(self, state, scope, pos, paramTypes, x, y, c00, c01, c10, c11,
d00, d01, d10, d11):
length = len(x)
if len(y) != length:
raise PFARuntimeException("dimensions of vectors do not match",
self.errcodeBase + 0, self.name, pos)
a00, a01, a10, a11 = self.countPairs(x, y)
return div((c11 * a11 + c10 * a10 + c01 * a01 + c00 * a00),
(d11 * a11 + d10 * a10 + d01 * a01 + d00 * a00))
def interpolateMulti(datum, one, two, code, fcnName, pos):
onex = one["x"]
twox = two["x"]
oney = one["to"]
twoy = two["to"]
unitless = div((datum - onex), (twox - onex))
if len(oney) != len(twoy):
raise PFARuntimeException("inconsistent dimensionality", code, fcnName, pos)
return [(1.0 - unitless)*oney[i] + unitless*twoy[i] for i in xrange(len(oney))]
def __call__(self, state, scope, pos, paramTypes, x, meanVariance, *args):
mean = meanVariance["mean"]
variance = meanVariance["variance"]
if variance >= 0.0:
sigma = math.sqrt(variance)
else:
sigma = float("nan")
if len(args) == 0:
return div(x - mean, sigma)
else:
unbiased, = args
count = meanVariance["count"]
if count / (count - 1.0) >= 0.0:
correction = math.sqrt((count) / (count - 1.0))
else:
correction = float("nan")
if unbiased:
return div(x - mean, sigma) * correction
else:
return div(x - mean, sigma)
def __call__(self, state, scope, pos, paramTypes, x, meanVariance, *args):
mean = meanVariance["mean"]
variance = meanVariance["variance"]
if variance >= 0.0:
sigma = math.sqrt(variance)
else:
sigma = float("nan")
if len(args) == 0:
return div(x - mean, sigma)
else:
unbiased, = args
count = meanVariance["count"]
if count / (count - 1.0) >= 0.0:
correction = math.sqrt((count) / (count - 1.0))
else:
correction = float("nan")
if unbiased:
return div(x - mean, sigma) * correction
else:
return div(x - mean, sigma)
def calculate(self, x, size):
symm = (x + x.T) * 0.5
evals, evects = np().linalg.eig(symm)
evects = np().array(evects)
evects2 = [evects[:,i] * (-1.0 if evects[0,i] < 0.0 else 1.0) for i in range(size)]
eigvalm2 = [div(1.0, math.sqrt(abs(ei))) for ei in evals]
order = np().argsort(eigvalm2)
out = np().empty((size, size), dtype=np().double)
for i in range(size):
for j in range(size):
out[i,j] = evects2[order[i]][j] * eigvalm2[order[i]]
return out
def calculate(self, x, size):
symm = (x + x.T) * 0.5
evals, evects = np().linalg.eig(symm)
evects = np().array(evects)
evects2 = [evects[:,i] * (-1.0 if evects[0,i] < 0.0 else 1.0) for i in xrange(size)]
eigvalm2 = [div(1.0, math.sqrt(abs(ei))) for ei in evals]
order = np().argsort(eigvalm2)
out = np().empty((size, size), dtype=np().double)
for i in xrange(size):
for j in xrange(size):
out[i,j] = evects2[order[i]][j] * eigvalm2[order[i]]
return out
def __call__(self, state, scope, paramTypes, data, cluster, weight):
if len(data) == 0:
raise PFARuntimeException("no data")
dimension = len(data[0])
summ = [0.0 for i in xrange(dimension)]
for vec in data:
if len(vec) != dimension:
raise PFARuntimeException("dimensions of vectors do not match")
for i in xrange(dimension):
summ[i] += vec[i]
vec = cluster["center"]
if len(vec) != dimension:
raise PFARuntimeException("dimensions of vectors do not match")
for i in xrange(dimension):
summ[i] += weight * vec[i]
denom = len(data) + weight
for i in xrange(dimension):
summ[i] = div(summ[i], denom)
return dict(cluster, center=summ)
def __call__(self, state, scope, pos, paramTypes, x, w, theState,
windowSize, level):
if windowSize < 2:
raise PFARuntimeException("windowSize must be at least 2",
self.errcodeBase + 0, self.name, pos)
if len(theState) == 0:
out = {}
paramType = state.parser.getAvroType(paramTypes[2])
for field in self._getRecord(paramType).fields:
if field.name == "x":
out["x"] = x
elif field.name == "w":
out["w"] = w
elif field.name == "count":
out["count"] = w
elif field.name == "mean":
out["mean"] = x
elif field.name == "variance":
out["variance"] = 0.0
else:
raise PFARuntimeException(
"cannot initialize unrecognized fields",
self.errcodeBase + 1, self.name, pos)
return [out]
elif len(theState) >= windowSize:
record = theState[-1]
splitAt = len(theState) - windowSize + 1
remove, keep = theState[:splitAt], theState[splitAt:]
oldx = [xi["x"] for xi in remove]
oldw = [-xi["w"] for xi in remove]
originalCount = record["count"]
count = originalCount + w
count2 = count + sum(oldw)
if level == 0:
return keep + [dict(record, x=x, w=w, count=count2)]
else:
mean = record["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
accumulatedCount = count
varianceCorrection = 0.0
for ox, ow in zip(oldx, oldw):
accumulatedCount += ow
delta2 = ox - mean
shift2 = div(delta2 * ow, accumulatedCount)
mean += shift2
varianceCorrection += (accumulatedCount -
ow) * delta2 * shift2
if level == 1:
return keep + [
dict(record, x=x, w=w, count=count2, mean=mean)
]
else:
varianceTimesCount = record["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
varianceTimesCount += varianceCorrection
return keep + [
dict(record,
x=x,
w=w,
count=count2,
mean=mean,
variance=div(varianceTimesCount, count2))
]
else:
record = theState[-1]
originalCount = record["count"]
count = originalCount + w
if level == 0:
return theState + [dict(record, x=x, w=w, count=count)]
else:
mean = record["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
if level == 1:
return theState + [
dict(record, x=x, w=w, count=count, mean=mean)
]
else:
varianceTimesCount = record["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
return theState + [
dict(record,
x=x,
w=w,
count=count,
mean=mean,
variance=div(varianceTimesCount, count))
]
def __call__(self, state, scope, pos, paramTypes, x, w, theState):
size = len(x)
if size < 2:
raise PFARuntimeException("too few components",
self.errcodeBase + 1, self.name, pos)
if paramTypes[0]["type"] == "map":
oldCount = theState["count"]
oldMean = theState["mean"]
oldCovariance = theState["covariance"]
countKeys = set(oldCount.keys()).union(
reduce(lambda a, b: a.union(b),
(set(v.keys()) for v in list(oldCount.values())),
set()))
covarKeys = set(oldCovariance.keys()).union(
reduce(lambda a, b: a.union(b),
(set(v.keys()) for v in list(oldCovariance.values())),
set()))
keys = set(x.keys()).union(countKeys).union(covarKeys)
newCount = {}
for i in keys:
row = {}
for j in keys:
old = oldCount.get(i, {}).get(j, 0.0)
if (i in x) and (j in x):
row[j] = old + w
else:
row[j] = old
newCount[i] = row
newMean = {}
for i in keys:
old = oldMean.get(i, 0.0)
if i in x:
newMean[i] = old + div((x[i] - old) * w, newCount[i][i])
else:
newMean[i] = old
newCovariance = {}
for i in keys:
row = {}
for j in keys:
oldCov = oldCovariance.get(i, {}).get(j, 0.0)
if (i in x) and (j in x):
oldC = oldCount.get(i, {}).get(j, 0.0)
oldMi = oldMean.get(i, 0.0)
oldMj = oldMean.get(j, 0.0)
row[j] = div(((oldCov * oldC) + div(
(x[i] - oldMi) *
(x[j] - oldMj) * w * oldC, newCount[i][j])),
newCount[i][j])
else:
row[j] = oldCov
newCovariance[i] = row
return dict(theState,
count=newCount,
mean=newMean,
covariance=newCovariance)
else:
oldCount = theState["count"]
oldMean = theState["mean"]
oldCovariance = theState["covariance"]
if (size != len(oldMean) or size != len(oldCovariance)
or any(size != len(xi) for xi in oldCovariance)):
raise PFARuntimeException("unequal length arrays",
self.errcodeBase + 2, self.name, pos)
newCount = oldCount + w
newMean = [
oldm + div((x[i] - oldm) * w, newCount)
for i, oldm in enumerate(oldMean)
]
newCovariance = []
for i in range(size):
row = []
for j in range(size):
row.append(
div((oldCovariance[i][j] * oldCount) + div(
(x[i] - oldMean[i]) *
(x[j] - oldMean[j]) * w * oldCount, newCount),
newCount))
newCovariance.append(row)
return dict(theState,
count=newCount,
mean=newMean,
covariance=newCovariance)
def __call__(self, state, scope, pos, paramTypes, x, y):
return div(x, y)
def __call__(self, state, scope, pos, paramTypes, x, y):
length = len(x)
if len(y) != length:
raise PFARuntimeException("dimensions of vectors do not match", self.errcodeBase + 0, self.name, pos)
a00, a01, a10, a11 = self.countPairs(x, y)
return div((a11 + a00), (a11 + 2*(a10 + a01) + a00))
def __call__(self, state, scope, pos, paramTypes, x, w, theState):
size = len(x)
if size < 2:
raise PFARuntimeException("too few components", self.errcodeBase + 1, self.name, pos)
if paramTypes[0]["type"] == "map":
oldCount = theState["count"]
oldMean = theState["mean"]
oldCovariance = theState["covariance"]
countKeys = set(oldCount.keys()).union(reduce(lambda a, b: a.union(b), (set(v.keys()) for v in oldCount.values()), set()))
covarKeys = set(oldCovariance.keys()).union(reduce(lambda a, b: a.union(b), (set(v.keys()) for v in oldCovariance.values()), set()))
keys = set(x.keys()).union(countKeys).union(covarKeys)
newCount = {}
for i in keys:
row = {}
for j in keys:
old = oldCount.get(i, {}).get(j, 0.0)
if (i in x) and (j in x):
row[j] = old + w
else:
row[j] = old
newCount[i] = row
newMean = {}
for i in keys:
old = oldMean.get(i, 0.0)
if i in x:
newMean[i] = old + div((x[i] - old) * w, newCount[i][i])
else:
newMean[i] = old
newCovariance = {}
for i in keys:
row = {}
for j in keys:
oldCov = oldCovariance.get(i, {}).get(j, 0.0)
if (i in x) and (j in x):
oldC = oldCount.get(i, {}).get(j, 0.0)
oldMi = oldMean.get(i, 0.0)
oldMj = oldMean.get(j, 0.0)
row[j] = div(((oldCov*oldC) + div((x[i] - oldMi) * (x[j] - oldMj) * w*oldC, newCount[i][j])), newCount[i][j])
else:
row[j] = oldCov
newCovariance[i] = row
return dict(theState, count=newCount, mean=newMean, covariance=newCovariance)
else:
oldCount = theState["count"]
oldMean = theState["mean"]
oldCovariance = theState["covariance"]
if (size != len(oldMean) or size != len(oldCovariance) or any(size != len(xi) for xi in oldCovariance)):
raise PFARuntimeException("unequal length arrays", self.errcodeBase + 2, self.name, pos)
newCount = oldCount + w
newMean = [oldm + div((x[i] - oldm) * w, newCount) for i, oldm in enumerate(oldMean)]
newCovariance = []
for i in xrange(size):
row = []
for j in xrange(size):
row.append(div((oldCovariance[i][j]*oldCount) + div((x[i] - oldMean[i]) * (x[j] - oldMean[j]) * w*oldCount, newCount), newCount))
newCovariance.append(row)
return dict(theState, count=newCount, mean=newMean, covariance=newCovariance)
def __call__(self, state, scope, pos, paramTypes, state_):
return div(state_["chi2"], state_["dof"])
def __call__(self, state, scope, pos, paramTypes, x, y, c00, c01, c10, c11, d00, d01, d10, d11):
length = len(x)
if len(y) != length:
raise PFARuntimeException("dimensions of vectors do not match", self.errcodeBase + 0, self.name, pos)
a00, a01, a10, a11 = self.countPairs(x, y)
return div((c11*a11 + c10*a10 + c01*a01 + c00*a00), (d11*a11 + d10*a10 + d01*a01 + d00*a00))
def __call__(self, state, scope, pos, paramTypes, x, y, sigma):
return math.exp(div(-math.log(2) * (x - y)**2, sigma**2))
def __call__(self, state, scope, pos, paramTypes, state_):
return div(state_["chi2"], state_["dof"])
def __call__(self, state, scope, paramTypes, x, w, theState, windowSize, level):
if windowSize < 2:
raise PFARuntimeException("windowSize must be at least 2")
if len(theState) == 0:
out = {}
paramType = state.parser.getAvroType(paramTypes[2])
for field in self._getRecord(paramType).fields:
if field.name == "x":
out["x"] = x
elif field.name == "w":
out["w"] = w
elif field.name == "count":
out["count"] = w
elif field.name == "mean":
out["mean"] = x
elif field.name == "variance":
out["variance"] = 0.0
else:
raise PFARuntimeException("cannot initialize unrecognized fields")
return [out]
elif len(theState) >= windowSize:
record = theState[-1]
splitAt = len(theState) - windowSize + 1
remove, keep = theState[:splitAt], theState[splitAt:]
oldx = [xi["x"] for xi in remove]
oldw = [-xi["w"] for xi in remove]
originalCount = record["count"]
count = originalCount + w
count2 = count + sum(oldw)
if level == 0:
return keep + [dict(record, x=x, w=w, count=count2)]
else:
mean = record["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
accumulatedCount = count
varianceCorrection = 0.0
for ox, ow in zip(oldx, oldw):
accumulatedCount += ow
delta2 = ox - mean
shift2 = div(delta2 * ow, accumulatedCount)
mean += shift2
varianceCorrection += (accumulatedCount - ow) * delta2 * shift2
if level == 1:
return keep + [dict(record, x=x, w=w, count=count2, mean=mean)]
else:
varianceTimesCount = record["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
varianceTimesCount += varianceCorrection
return keep + [dict(record, x=x, w=w, count=count2, mean=mean, variance=div(varianceTimesCount, count2))]
else:
record = theState[-1]
originalCount = record["count"]
count = originalCount + w
if level == 0:
return theState + [dict(record, x=x, w=w, count=count)]
else:
mean = record["mean"]
delta = x - mean
shift = div(delta * w, count)
mean += shift
if level == 1:
return theState + [dict(record, x=x, w=w, count=count, mean=mean)]
else:
varianceTimesCount = record["variance"] * originalCount
varianceTimesCount += originalCount * delta * shift
return theState + [dict(record, x=x, w=w, count=count, mean=mean, variance=(varianceTimesCount / count))]
def __call__(self, state, scope, pos, paramTypes, x, y, sigma):
return math.exp(div(-math.log(2) * (x - y)**2, sigma**2))
def __call__(self, state, scope, pos, paramTypes, x, y):
return div(x, y)
