/usr/share/pyshared/brian/tools/autodiff.py is in python-brian 1.3.1-1build1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 | '''
Automatic differentiation.
(C) R. Brette 2008
This file is released under the terms of the BSD licence
--------------------------------------------------------
Method: forward accumulation by operator overloading.
Use::
df=differentiate(f) # where f is a single-variable function
y=differentiate(f,x0) # differentiate at x=x0
y=differentiate(f,x0,order=2) # second-order derivative at x=x0
y=differentiate(f,(a,b),order=(1,2)) # d2f/dxdy(a,b)
y=differentiate(f,order=(1,2)) # d2f/dxdy
a0,a1,a2=taylor_series(f,x0,order=2) # 2nd order Taylor series of f at x0
J=gradient(f,(a,b)) # gradient of f at (a,b) (returns a list)
H=hessian(f,(a,b)) # hessian of f at (a,b) (returns a numpy array)
TODO:
* replace x by x0
* merge HigherDifferentiable in Differentiable?
* simplify code
* gradient, taylor_series, hessian: x0 should be optional
* iadd, etc. (more operators)
* NonDifferentiableException and maybe use this on conditions (when cmp=0)
* Differential operators: Laplacian...
* gradient, taylor_series, hessian: x0 should be optional
* test with units, functions Rp->something, alternative differential operators,
differential (f:Rp->Rn), complex numbers, arithmetic derivative
* remove most dependency with numpy (except Hessian)
* implicitly defined functions
If several partial derivatives of the same function need to be
calculated, it is faster to use the lower level object Differentiable.
Example::
y=f(Differentiable('x',3),Differentiable('y',2))
will return a Differentiable object with y.val=f(3,2)
and y.diff={'x': df/dx(3,2), 'y': df/dy(3,2)}
'''
from numpy import exp, sin, cos, log, zeros # todo: look in frame instead?
import types
__all__ = ['differentiate', 'Differentiable', 'taylor_series', 'gradient', 'hessian']
def differentiate(f, x=None, order=1):
'''
Calculate the derivative of f at point x.
Higher-order derivatives are possible.
Ex.: differentiate(lambda x:3*x*x+2,x=2,order=2)
(Returns: 6.0)
Partial derivatives:
1) Give a tuple with the order.
Ex.: differentiate(lambda x,y:x*y+2*y+1,x=(1,2),order=(0,1))
(Returns: 3.0 (partial derivative d/dy) )
2) Give a dictionnary with the order:
Ex.: differentiate(lambda x,y:x*y+2*y+1,x=(1,2),order={'y':1})
'''
if x is None:
return lambda x:differentiate(f, x, order)
if (type(order) == types.ListType) or (type(order) == types.TupleType): # several variables
# Build the list of arguments
n = sum(order) # total order
args = [HigherDifferentiable(i, x[i], n) for i in range(len(x))]
y = f(*args)
for i in xrange(len(x)):
for _ in xrange(order[i]):
if isinstance(y, Differentiable) and (i in y.diff):
y = y.diff[i]
else: # constant
return 0.
return y
elif (type(order) == types.DictType): # named variables
# Build the list of arguments
n = sum(order.itervalues())
vars = list(f.func_code.co_varnames)
args = []
args.extend(x)
for name in order.iterkeys():
i = vars.index(name)
args[i] = HigherDifferentiable(name, x[i], n)
y = f(*args)
for name, n in order.iteritems():
for j in range(n):
if isinstance(y, Differentiable) and (name in y.diff):
y = y.diff[name]
else: # constant
return 0.
return y
elif order == 0:
return f(x)
elif order == 1:
y = f(Differentiable('x', x))
if isinstance(y, Differentiable) and ('x' in y.diff):
return y.diff['x']
else: # constant
return 0.
else:
return differentiate(lambda y:differentiate(f, y, order=order - 1), x)
def taylor_series(f, x, order=1):
'''
Returns the list of coefficients of the Taylor Series of f
at x (scalar).
'''
y = f(HigherDifferentiable('x', x, order))
series = []
n = 1.
for i in range(order + 1):
# Value
val = y
while isinstance(val, Differentiable):
val = val.val
series.append(val / n)
n *= (i + 1)
# Next order
if isinstance(y, Differentiable) and ('x' in y.diff):
y = y.diff['x']
else:
y = 0.
return series
def gradient(f, x):
'''
Gradient of f at x (= tuple).
Result = list.
'''
args = [Differentiable(i, val=x[i]) for i in range(len(x))]
y = f(*args)
if isinstance(y, Differentiable):
result = []
for i in range(len(x)):
if i in y.diff:
result.append(y.diff[i])
else:
result.append(0.)
return result
else: # constant
return [0.] * len(x)
def hessian(f, x):
'''
Hessian of f at x (= tuple).
Result = array.
N.B.: not working with vectors.
'''
args = [HigherDifferentiable(i, x[i], 2) for i in range(len(x))]
y = f(*args)
result = zeros((len(x), len(x)))
if isinstance(y, Differentiable): # non-zero Hessian
for i in range(len(x)):
if i in y.diff and isinstance(y.diff[i], Differentiable):
for j in range(len(x)):
if j in y.diff[i].diff:
result[i, j] = y.diff[i].diff[j]
return result
class Differentiable(object):
'''
A differentiable variable.
Implemented operations: +,-,*,[],len,exp,sin,cos,/,abs,sqrt,comparisons,**
'''
def __init__(self, name=None, val=0.):
'''
Initializes a variable and its derivative.
If the name is None, then it is a constant.
'''
self.val = val # value
if name == None: # constant
self.diff = {} # derivative
else:
self.diff = {name:1.} # or jacobian? (eye)
def sqrt(self):
return self ** .5
def __cmp__(self, x):
if isinstance(x, Differentiable):
return cmp(self.val, x.val)
else:
return cmp(self.val, x)
def __abs__(self):
# NOT WORKING WITH VECTORS
if self.val == 0:
# TODO: specific exception
raise ZeroDivisionError, "x:abs(x) is not differentiable at x=0."
elif self.val > 0:
return self
else:
return - self
def __add__(self, y):
# VECTOR-READY
if isinstance(y, Differentiable):
zdict = {}
zdict.update(self.diff)
zdict.update(y.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in y.diff):
zdict[key] = self.diff[key] + y.diff[key]
z = Differentiable()
z.val = self.val + y.val
z.diff = zdict
return z
else: # Adding a constant
z = Differentiable(val=self.val + y)
z.diff = self.diff
return z
def __radd__(self, x):
# VECTOR-READY
if isinstance(x, Differentiable):
zdict = {}
zdict.update(self.diff)
zdict.update(x.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in x.diff):
zdict[key] = x.diff[key] + self.diff[key]
z = Differentiable()
z.val = x.val + self.val
z.diff = zdict
return z
else: # Adding a constant
z = Differentiable(val=x + self.val)
z.diff = self.diff
return z
def __mul__(self, y):
# !Not working with vectors!
if isinstance(y, Differentiable):
#TODO: with vectors
zdict = {}
zdict.update(self.diff)
zdict.update(y.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in y.diff):
zdict[key] = self.diff[key] * y.val + self.val * y.diff[key]
elif (key in self.diff):
zdict[key] = self.diff[key] * y.val
else:
zdict[key] = self.val * y.diff[key]
z = Differentiable()
z.val = self.val * y.val
z.diff = zdict
return z
else: # Multiplying by a constant
z = Differentiable(val=self.val * y)
for key in self.diff:
z.diff[key] = self.diff[key] * y
return z
def __rmul__(self, x):
if isinstance(x, Differentiable):
#TODO: with vectors
zdict = {}
zdict.update(self.diff)
zdict.update(x.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in x.diff):
zdict[key] = x.val * self.diff[key] + x.diff[key] * self.val
elif (key in self.diff):
zdict[key] = x.val * self.diff[key]
else:
zdict[key] = x.diff[key] * self.val
z = Differentiable()
z.val = x.val * self.val
z.diff = zdict
return z
else: # Multiplying by a constant
z = Differentiable(val=x * self.val)
for key in self.diff:
z.diff[key] = x * self.diff[key]
return z
def __div__(self, x):
return self * (x ** -1)
def __rdiv__(self, x):
return x * (self ** -1)
def __sub__(self, y):
# VECTOR-READY
if isinstance(y, Differentiable):
zdict = {}
zdict.update(self.diff)
zdict.update(y.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in y.diff):
zdict[key] = self.diff[key] - y.diff[key]
z = Differentiable()
z.val = self.val - y.val
z.diff = zdict
return z
else: # Subtracting a constant
z = Differentiable(val=self.val - y)
z.diff = self.diff
return z
def __rsub__(self, x):
# VECTOR-READY
if isinstance(x, Differentiable):
zdict = {}
zdict.update(self.diff)
zdict.update(x.diff)
for key in zdict.iterkeys():
if (key in self.diff) and (key in x.diff):
zdict[key] = x.diff[key] - self.diff[key]
z = Differentiable()
z.val = x.val - self.val
z.diff = zdict
return z
else: # Subtracting a constant
z = Differentiable(val=x - self.val)
for key in self.diff:
z.diff[key] = -self.diff[key]
return z
def __neg__(self):
# VECTOR-READY
z = Differentiable(val= -self.val)
for key in self.diff:
z.diff[key] = -self.diff[key]
return z
def __pow__(self, x):
# NOT WORKING WITH VECTORS
if isinstance(x, Differentiable):
return exp(x * log(self))
elif x == 0:
z = Differentiable(val=self.val ** x)
for key in self.diff:
z.diff[key] = 0 * self.diff[key]
return z
else:
z = Differentiable(val=self.val ** x)
for key in self.diff:
z.diff[key] = x * self.val ** (x - 1) * self.diff[key]
return z
def __rpow__(self, x):
# NOT WORKING WITH VECTORS
return exp(self * log(x))
def __getitem__(self, i):
z = Differentiable(val=self.val[i])
z.diff = {}
for key in self.diff.iterkeys():
z.diff[key] = self.diff[key][i, :]
return z
def __len__(self):
return len(self.val)
def __str__(self):
s = str(self.val)
for key, value in self.diff.iteritems():
s += ' ; d/d' + key + '=' + str(value)
return s
def exp(self):
# !! Not working for vectors !!
zdict = {}
zdict.update(self.diff)
for key in zdict.iterkeys():
zdict[key] = exp(self.val) * self.diff[key]
z = Differentiable(val=exp(self.val))
z.diff = zdict
return z
def cos(self):
# !! Not working for vectors !!
zdict = {}
zdict.update(self.diff)
for key in zdict.iterkeys():
zdict[key] = -sin(self.val) * self.diff[key]
z = Differentiable(val=cos(self.val))
z.diff = zdict
return z
def sin(self):
# !! Not working for vectors !!
zdict = {}
zdict.update(self.diff)
for key in zdict.iterkeys():
zdict[key] = cos(self.val) * self.diff[key]
z = Differentiable(val=sin(self.val))
z.diff = zdict
return z
def log(self):
# !! Not working for vectors !!
zdict = {}
zdict.update(self.diff)
for key in zdict.iterkeys():
zdict[key] = (1. / (self.val)) * self.diff[key]
z = Differentiable(val=log(self.val))
z.diff = zdict
return z
class HigherDifferentiable(Differentiable):
'''
A differentiable variable with order n.
'''
def __init__(self, name=None, val=0., order=1):
Differentiable.__init__(self, name, val)
if order > 1:
self.val = HigherDifferentiable(name, val, order - 1)
if __name__ == '__main__':
print "Derivative:"
print "> differentiate(lambda x:3*x*x+2,x=2,order=2)"
print differentiate(lambda x:3 * x * x + 2, x=2, order=2)
print "> differentiate(lambda x:3*x*x+2,order=2)(2)"
print differentiate(lambda x:3 * x * x + 2, order=2)(2)
print "Partial derivative:"
print "> differentiate(lambda x,y:x*y+2*y+1,x=(1,2),order=(0,1))"
print differentiate(lambda x, y:x * y + 2 * y + 1, x=(1, 2), order=(0, 1))
|