Develop Reference

Substituting matrix for a scalar in SymPy

Here's what I'm doing in a SymPy session:
from sympy import *
xi1,xi2,xi3 = symbols('xi_1,xi_2,xi_3')
N1 = 1-xi1-xi2-xi3
N2 = xi3
N3 = xi1
N4 = xi2
x1,x2,x3,x4 = symbols('x_1, x_2, x_3, x_4')
x = N1*x1+N2*x2+N3*x3+N2*x4
subdict = {x1:Matrix([0.025,1.0,0.0]), x2 : Matrix([0,1,0]), x3:Matrix([0, 0.975, 0]), x4:Matrix([0,0.975,0.025])}
x.subs(subdict)
test.subs({xi1:1, xi2:0,xi3:0})
To me we are simply multiplying some scalars with some vectors then adding them up. However SymPy would disagree and throws a ginormous error for which the last line is:
TypeError: cannot add <class 'sympy.matrices.immutable.ImmutableDenseMatrix'> and <class 'sympy.core.numbers.Zero'>
Why is this a problem? Is there a workaround for what I am trying to do?

I suspect that what is happening is that before the matrix is substituted you substitute a 0 which makes 0*matrix_symbol = 0 instead of a matrix of zeros. Terms that end up being matrices cannot be added to 0 and thus the error. My attempts at using the simultaneous flag or xreplace instead of subs give the same result (on sympy.live.org). Then I tried to do the substitutions in reversed order, passing them as a list with the matrices first. Still didn't work. It looks like subs assumes that 0*foo is 0. An issue at sympy issues should be raised if there is not already an existing issue.
The workaround is to do the scalar substitutions first, allowing the zero terms to disappear. Then do a subs with the matrices. So this will require 2 calls to subs.
A true, hackish workaround for doing substitution with 0 is this:
def remul(m):
rv = 1
for i in m.args:
rv *= i
return rv
expr = x*y
mat = expr.subs(x, randMatrix(2)) # replace x with matrix
expr = mat.replace( # replace y with scalar 0
lambda m: m.is_Mul,
lambda m: remul(Mul(*[i.subs(y, 0) for i in m.args], evaluate=False)))

How to Define a Piecewise Function in Python with Multiple Variables

I am trying to develop a plot for my helioseismology class and the question had provided a piecewise function describing the dynamics of the "fluids" in a star as if it is one thing its this and if its another its that. I am receiving over and over again this 'Mul' object cannot be interpreted as an integer but I am working with numbers in the reals not just the integer set. I do not know how to get around this and need guidance. The code is as follows.
import sympy as sy
from sympy import *
from sympy.physics.units import Unit
import numpy as np
import sys
import math
import scipy as sp
from scipy import special
phi = Symbol('phi', Variable = True)
x = Symbol('x', Variable = True, Real = True)
t = Symbol('t', Variable = True, Real = True)
xi = Symbol('xi', Function = True)
Solar_Radius = Symbol('R', Constant = True, unit = "meters")
Sound_Speed = Symbol('c', Constant = True, unit = "meters per second", Real = True)
gamma = Symbol('gamma', Constant = True)
gravity = Symbol('g', Constant = True, unit = "meters per second per second")
Solar_Radius = 6.963 * 10 ** 6
gamma = 5/3
g = 274.8265625336
gas_constant = 8201.25
c = 8.1 * 10 ** 3
for t in range(0,x/c):
xi[x,t] = 0
for t in range(x/c,00):
xi[x,t] = (1/2)*sy.exp(gamma*g*x/(2*c**2))*mpmath.besselj(0, (gamma*g/(2*c)*sy.sqrt(t**2 - ((x/c)**2))),derivative = 0)
Full Traceback:
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-50-3506376f1686> in <module>()
----> 1 for t in range(0,x/c):
2 xi[x,t] = 0
3 for t in range(x/c,00):
4 xi[x,t] = (1/2)*sy.exp(gamma*g*x/(2*c**2))*mpmath.besselj(0, (gamma*g/(2*c)*sy.sqrt(t**2 - ((x/c)**2))),derivative = 0)
TypeError: 'Mul' object cannot be interpreted as an integer

There are quite a few issues here:
None of the keyword arguments (Constant, Variable, unit, Real) that you are passing to Symbol are things that are recognized by SymPy. The only one that is close is real, which should be lowercase (like Symbol('x', real=True)). The rest do nothing. If you want units, you should use the SymPy units module in sympy.physics.units. There is no need to specify if a symbol is constant or variable.
You have redefined Solar_Radius and gamma as numbers. That means that the Symbol definitions for those variables are pointless.
If you are using Python 2, make sure to include from __future__ import division at the top of the file, or else things like 1/2 and 5/3 will be truncated with integer division (this isn't an issue in Python 3).
range(0, x/c) doesn't make sense. range creates a list of numbers, like range(0, 3) -> [0, 1, 2]. But x/c is not a number, it's a symbolic expression.
Additionally, xi[x, t] = ... doesn't make sense. xi is a Symbol, which doesn't allow indexing and certainly doesn't allow assignment.
Don't mix numeric (math, mpmath, numpy, scipy) functions with SymPy functions. They won't work with symbolic expressions. You should use only SymPy functions. If you create a symbolic expression and want to convert it to a numeric one (e.g., for plotting), use lambdify.
What you want here is Piecewise. The syntax is Piecewise((expr, cond), (expr, cond), ..., (expr, True)), where expr is an expression that is used when cond is true ((expr, True) is the "otherwise" condition).
For your example, I believe you want
expr = Piecewise((0, t < x/c), (sy.exp(gamma*g*x/(2*c**2))*sy.besselj(0, (gamma*g/(2*c)*sy.sqrt(t**2 - (x/c)**2)))/2, t >= x/c))
If you want to turn this into a numeric function in x and t, use
xi = lambdify((x, t), expr)

Writing a function for x * sin(3/x) in python

I have to write a function, s(x) = x * sin(3/x) in python that is capable of taking single values or vectors/arrays, but I'm having a little trouble handling the cases when x is zero (or has an element that's zero). This is what I have so far:
def s(x):
result = zeros(size(x))
for a in range(0,size(x)):
if (x[a] == 0):
result[a] = 0
else:
result[a] = float(x[a] * sin(3.0/x[a]))
return result
Which...doesn't work for x = 0. And it's kinda messy. Even worse, I'm unable to use sympy's integrate function on it, or use it in my own simpson/trapezoidal rule code. Any ideas?
When I use integrate() on this function, I get the following error message: "Symbol" object does not support indexing.

This takes about 30 seconds per integrate call:
import sympy as sp
x = sp.Symbol('x')
int2 = sp.integrate(x*sp.sin(3./x),(x,0.000001,2)).evalf(8)
print int2
int1 = sp.integrate(x*sp.sin(3./x),(x,0,2)).evalf(8)
print int1
The results are:
1.0996940
-4.5*Si(zoo) + 8.1682775
Clearly you want to start the integration from a small positive number to avoid the problem at x = 0.
You can also assign x*sin(3./x) to a variable, e.g.:
s = x*sin(3./x)
int1 = sp.integrate(s, (x, 0.00001, 2))
My original answer using scipy to compute the integral:
import scipy.integrate
import math
def s(x):
if abs(x) < 0.00001:
return 0
else:
return x*math.sin(3.0/x)
s_exact = scipy.integrate.quad(s, 0, 2)
print s_exact
See the scipy docs for more integration options.

If you want to use SymPy's integrate, you need a symbolic function. A wrong value at a point doesn't really matter for integration (at least mathematically), so you shouldn't worry about it.
It seems there is a bug in SymPy that gives an answer in terms of zoo at 0, because it isn't using limit correctly. You'll need to compute the limits manually. For example, the integral from 0 to 1:
In [14]: res = integrate(x*sin(3/x), x)
In [15]: ans = limit(res, x, 1) - limit(res, x, 0)
In [16]: ans
Out[16]:
9⋅π 3⋅cos(3) sin(3) 9⋅Si(3)
- ─── + ──────── + ────── + ───────
4 2 2 2
In [17]: ans.evalf()
Out[17]: -0.164075835450162

How can I avoid value errors when using numpy.random.multinomial?

When I use this random generator: numpy.random.multinomial, I keep getting:
ValueError: sum(pvals[:-1]) > 1.0
I am always passing the output of this softmax function:
def softmax(w, t = 1.0):
e = numpy.exp(numpy.array(w) / t)
dist = e / np.sum(e)
return dist
except now that I am getting this error, I also added this for the parameter (pvals):
while numpy.sum(pvals) > 1:
pvals /= (1+1e-5)
but that didn't solve it. What is the right way to make sure I avoid this error?
EDIT: here is function that includes this code
def get_MDN_prediction(vec):
coeffs = vec[::3]
means = vec[1::3]
stds = np.log(1+np.exp(vec[2::3]))
stds = np.maximum(stds, min_std)
coe = softmax(coeffs)
while np.sum(coe) > 1-1e-9:
coe /= (1+1e-5)
coeff = unhot(np.random.multinomial(1, coe))
return np.random.normal(means[coeff], stds[coeff])

I also encountered this problem during my language modelling work.
The root of this problem rises from numpy's implicit data casting: the output of my sorfmax() is in float32 type, however, numpy.random.multinomial() will cast the pval into float64 type IMPLICITLY. This data type casting would cause pval.sum() exceed 1.0 sometimes due to numerical rounding.
This issue is recognized and posted here

I know the question is old but since I faced the same problem just now, it seems to me it's still valid. Here's the solution I've found for it:
a = np.asarray(a).astype('float64')
a = a / np.sum(a)
b = np.random.multinomial(1, a, 1)
I've made the important part bold. If you omit that part the problem you've mentioned will happen from time to time. But if you change the type of array into float64, it will never happen.

Something that few people noticed: a robust version of the softmax can be easily obtained by removing the logsumexp from the values:
from scipy.misc import logsumexp
def log_softmax(vec):
return vec - logsumexp(vec)
def softmax(vec):
return np.exp(log_softmax(vec))
Just check it:
print(softmax(np.array([1.0, 0.0, -1.0, 1.1])))
Simple, isn't it?

The softmax implementation I was using is not stable enough for the values I was using it with. As a result, sometimes the output has a sum greater than 1 (e.g. 1.0000024...).
This case should be handled by the while loop. But sometimes the output contains NaNs, in which case the loop is never triggered, and the error persists.
Also, numpy.random.multinomial doesn't raise an error if it sees a NaN.
Here is what I'm using right now, instead:
def softmax(vec):
vec -= min(A(vec))
if max(vec) > 700:
a = np.argsort(vec)
aa = np.argsort(a)
vec = vec[a]
i = 0
while max(vec) > 700:
i += 1
vec -= vec[i]
vec = vec[aa]
e = np.exp(vec)
return e/np.sum(e)
def sample_multinomial(w):
"""
Sample multinomial distribution with parameters given by softmax of w
Returns an int
"""
p = softmax(w)
x = np.random.uniform(0,1)
for i,v in enumerate(np.cumsum(p)):
if x < v: return i
return len(p)-1 # shouldn't happen...

Sympy: working with equalities manually

I'm currently doing a maths course where my aim is to understand the concepts and process rather than crunch through problem sets as fast as possible. When solving equations, I'd like to be able to poke at them myself rather than have them solved for me.
Let's say we have the very simple equation z + 1 = 4- if I were to solve this myself, I would obviously subtract 1 from both sides, but I can't figure out if sympy provides a simple way to do this. At the moment the best solution I can come up with is:
from sympy import *
z = symbols('z')
eq1 = Eq(z + 1, 4)
Eq(eq1.lhs - 1, eq1.rhs - 1)
# Output:
# z == 3
Where the more obvious expression eq1 - 1 only subtracts from the left-hand side. How can I use sympy to work through equalities step-by-step like this (i.e. without getting the solve() method to just given me the answer)? Any pointers to the manipulations that are actually possible with sympy equalities would be appreciated.

There is a "do" method and discussion at https://github.com/sympy/sympy/issues/5031#issuecomment-36996878 that would allow you to "do" operations to both sides of an Equality. It's not been accepted as an addition to SymPy but it is a simple add-on that you can use. It is pasted here for convenience:
def do(self, e, i=None, doit=False):
"""Return a new Eq using function given or a model
model expression in which a variable represents each
side of the expression.
Examples
========
>>> from sympy import Eq
>>> from sympy.abc import i, x, y, z
>>> eq = Eq(x, y)
When the argument passed is an expression with one
free symbol that symbol is used to indicate a "side"
in the Eq and an Eq will be returned with the sides
from self replaced in that expression. For example, to
add 2 to both sides:
>>> eq.do(i + 2)
Eq(x + 2, y + 2)
To add x to both sides:
>>> eq.do(i + x)
Eq(2*x, x + y)
In the preceding it was actually ambiguous whether x or i
was to be added but the rule is that any symbol that are
already in the expression are not to be interpreted as the
dummy variable. If we try to add z to each side, however, an
error is raised because now it is unclear whether i or z is being
added:
>>> eq.do(i + z)
Traceback (most recent call last):
...
ValueError: not sure what symbol is being used to represent a side
The ambiguity must be resolved by indicating with another parameter
which is the dummy variable representing a side:
>>> eq.do(i + z, i)
Eq(x + z, y + z)
Alternatively, if only one Dummy symbol appears in the expression then
it will be automatically used to represent a side of the Eq.
>>> eq.do(2*Dummy() + z)
Eq(2*x + z, 2*y + z)
Operations like differentiation must be passed as a
lambda:
>>> Eq(x, y).do(lambda i: i.diff(x))
Eq(1, 0)
Because doit=False by default, the result is not evaluated. to
evaluate it, either use the doit method or pass doit=True.
>>> _.doit == Eq(x, y).do(lambda i: i.diff(x), doit=True)
True
"""
if not isinstance(e, (FunctionClass, Lambda, type(lambda:1))):
e = S(e)
imaybe = e.free_symbols - self.free_symbols
if not imaybe:
raise ValueError('expecting a symbol')
if imaybe and i and i not in imaybe:
raise ValueError('indicated i not in given expression')
if len(imaybe) != 1 and not i:
d = [i for i in imaybe if isinstance(i, Dummy)]
if len(d) != 1:
raise ValueError(
'not sure what symbol is being used to represent a side')
i = set(d)
else:
i = imaybe
i = i.pop()
f = lambda side: e.subs(i, side)
else:
f = e
return self.func(*[f(side) for side in self.args], evaluate=doit)
from sympy.core.relational import Equality
Equality.do = do