def __init__(self):
x, y, z = scalars("xyz")
e = at.sigmoid((x + y + z)**2)
op = OpFromGraph([x, y, z], [e])
e2 = op(x, y, z)
self.inputs = [x, y, z]
self.outputs = [e2]
def invlogit(x, eps=None):
"""The inverse of the logit function, 1 / (1 + exp(-x))."""
if eps is not None:
warnings.warn(
"pymc.math.invlogit no longer supports the ``eps`` argument and it will be ignored.",
DeprecationWarning,
stacklevel=2,
)
return at.sigmoid(x)
def backward(self, rv_var, rv_value):
a, b = self.param_extract_fn(rv_var)
if a is not None and b is not None:
sigmoid_x = at.sigmoid(rv_value)
return sigmoid_x * b + (1 - sigmoid_x) * a
elif a is not None:
return at.exp(rv_value) + a
elif b is not None:
return b - at.exp(rv_value)
else:
return rv_value
def __init__(self, nfeatures=100, noutputs=10, nhiddens=50, rng=None):
if rng is None:
rng = 0
if isinstance(rng, int):
rng = np.random.default_rng(rng)
self.rng = rng
self.nfeatures = nfeatures
self.noutputs = noutputs
self.nhiddens = nhiddens
x = dmatrix("x")
wh = shared(self.rng.normal(0, 1, (nfeatures, nhiddens)), borrow=True)
bh = shared(np.zeros(nhiddens), borrow=True)
h = at.sigmoid(at.dot(x, wh) + bh)
wy = shared(self.rng.normal(0, 1, (nhiddens, noutputs)))
by = shared(np.zeros(noutputs), borrow=True)
y = at.nnet.softmax(at.dot(h, wy) + by)
self.inputs = [x]
self.outputs = [y]
