def define(name, expr):
"""
Take an expression of 't' (possibly complicated)
and make it a '%s(t)' % name, such that
when it evaluates it has the right values.
Parameters
----------
expr : sympy expression, with only 't' as a Symbol
name : str
Returns
-------
nexpr: sympy expression
Examples
--------
>>> t = Term('t')
>>> expr = t**2 + 3*t
>>> print expr
3*t + t**2
>>> newexpr = define('f', expr)
>>> print newexpr
f(t)
>>> import aliased
>>> f = aliased.lambdify(t, newexpr)
>>> f(4)
28
>>> 3*4+4**2
28
>>>
"""
v = vectorize(expr)
return aliased_function(name, v)(Term('t'))
def _getint(f, dt=0.02, t=50):
lf = vectorize(f)
tt = np.arange(dt,t+dt,dt)
return lf(tt).sum() * dt
def convolve_functions(fn1, fn2, interval, dt, padding_f=0.1, name=None):
"""
Convolve fn1 with fn2.
Parameters
----------
fn1 : sympy expr
An expression that is a function of t only.
fn2 : sympy expr
An expression that is a function of t only.
interval : [float, float]
The interval over which to convolve the two functions.
dt : float
Time step for discretization
padding_f : float
Padding added to the left and right in the convolution.
name : str
Name of the convolved function in the resulting expression.
Defaults to one created by linear_interp.
Returns
-------
f : sympy expr
An expression that is a function of t only.
>>> import sympy
>>> t = sympy.Symbol('t')
>>> # This is a square wave on [0,1]
>>> f1 = (t > 0) * (t < 1)
>>> # The convolution of with itself is a triangular wave on [0,2], peaking at 1 with height 1
>>> tri = convolve_functions(f1, f1, [0,2], 1.0e-03, name='conv')
>>> print tri
conv(t)
>>> ftri = vectorize(tri)
>>> x = np.linspace(0,2,11)
>>> y = ftri(x)
>>> # This is the resulting y-value (which seem to be numerically off by dt
>>> y
array([ -3.90255908e-16, 1.99000000e-01, 3.99000000e-01,
5.99000000e-01, 7.99000000e-01, 9.99000000e-01,
7.99000000e-01, 5.99000000e-01, 3.99000000e-01,
1.99000000e-01, 6.74679706e-16])
>>>
"""
max_interval, min_interval = max(interval), min(interval)
ltime = max_interval - min_interval
time = np.arange(min_interval, max_interval + padding_f * ltime, dt)
f1 = vectorize(fn1)
f2 = vectorize(fn2)
_fn1 = np.array(f1(time))
_fn2 = np.array(f2(time))
_fft1 = FFT.rfft(_fn1)
_fft2 = FFT.rfft(_fn2)
value = FFT.irfft(_fft1 * _fft2) * dt
_minshape = min(time.shape[0], value.shape[-1])
time = time[0:_minshape]
value = value[0:_minshape]
return linear_interp(time + min_interval, value, bounds_error=False, name=name)
coef = peak_location**(-alpha) * np.exp(peak_location / beta)
return coef * ((t >= 0) * (t+1.0e-14))**(alpha) * exp(-(t+1.0e-14)/beta)
# Glover canonical HRF models
# they are both Sympy objects
def _getint(f, dt=0.02, t=50):
lf = vectorize(f)
tt = np.arange(dt,t+dt,dt)
return lf(tt).sum() * dt
deft = DeferredVector('t')
_gexpr = gamma_params(5.4, 5.2) - 0.35 * gamma_params(10.8,7.35)
_gexpr = _gexpr / _getint(_gexpr)
_glover = vectorize(_gexpr)
glover = aliased_function('glover', _glover)
n = {}
glovert = vectorize(glover(deft))
# Derivative of Glover HRF
_dgexpr = _gexpr.diff(t)
dpos = Derivative((t >= 0), t)
_dgexpr = _dgexpr.subs(dpos, 0)
_dgexpr = _dgexpr / _getint(abs(_dgexpr))
_dglover = vectorize(_dgexpr)
dglover = aliased_function('dglover', _dglover)
dglovert = vectorize(dglover(deft))
del(_glover); del(_gexpr); del(dpos); del(_dgexpr); del(_dglover)
