概説微分積分の問題4.7と4.8の解答。手計算に飽きたので現実逃避。
4.7 は sympy を使う。理論的に解けるのはわかってるけども、手計算じゃもはや無理。
7番目の問題にやたら時間がかかったんだけど、辛かったのはどの辺だろ。セオリー的にはt = tan(x/2)と置いて置換積分して有理式に持ち込むんだろうけど、sympyの実装がどうなってるかは見てない。
import sympy as sym from sympy.abc import x print sym.integrate(x**2 / sym.sqrt(4 - x**2), x) print sym.integrate(x**2 / sym.sqrt(x**2 + 4), x) print sym.integrate(1 / (x**2 + 4)**5, x) print sym.integrate(1 / (x**2 + 2 * x * 10)**4, x) print sym.integrate(1 / sym.sin(x)**5, x) print sym.integrate(sym.log(x)**4, x) print sym.integrate(sym.sin(x) / (sym.cos(x)**2 + sym.sin(x) + 1), x) print sym.integrate(x * sym.sqrt((x - 1) * (x + 2)), x) print sym.integrate(1 / (sym.exp(x) + sym.exp(-x) + 1), x) print sym.integrate(sym.exp(x) * sym.sin(2 * x), x) print sym.integrate(x**2 * sym.exp(x) * sym.cos(x), x) print sym.integrate(x**2 * sym.sin(sym.sqrt(x)), x)
-x*(-x**2 + 4)**(1/2)/2 + 2*asin(x/2) x*(x**2 + 4)**(1/2)/2 - 2*asinh(x/2) (105*x**7 + 1540*x**5 + 8176*x**3 + 17856*x)/(98304*x**8 + 1572864*x**6 + 9437184*x**4 + 25165824*x**2 + 25165824) + 35*atan(x/2)/65536 -log(x)/64000000 + log(x + 20)/64000000 - (3*x**5 + 150*x**4 + 2200*x**3 + 6000*x**2 - 24000*x + 160000)/(9600000*x**6 + 576000000*x**5 + 11520000000*x**4 + 76800000000*x**3) (3*cos(x)**3 - 5*cos(x))/(8*cos(x)**4 - 16*cos(x)**2 + 8) + 3*log(cos(x) - 1)/16 - 3*log(cos(x) + 1)/16 x*log(x)**4 - 4*x*log(x)**3 + 12*x*log(x)**2 - 24*x*log(x) + 24*x -2*3**(1/2)*I*log(tan(x/2) - 1/2 - 3**(1/2)*I/2)*tan(x/2)/(9*tan(x/2) + 9) - 2*3**(1/2)*I*log(tan(x/2) - 1/2 - 3**(1/2)*I/2)/(9*tan(x/2) + 9) + 2*3**(1/2)*I*log(tan(x/2) - 1/2 + 3**(1/2)*I/2)*tan(x/2)/(9*tan(x/2) + 9) + 2*3**(1/2)*I*log(tan(x/2) - 1/2 + 3**(1/2)*I/2)/(9*tan(x/2) + 9) + 6/(9*tan(x/2) + 9) Integral(x*((x - 1)*(x + 2))**(1/2), x) 3**(1/2)*I*log(1/2 - 3**(1/2)*I/2 + exp(-x))/3 - 3**(1/2)*I*log(1/2 + 3**(1/2)*I/2 + exp(-x))/3 exp(x)*sin(2*x)/5 - 2*exp(x)*cos(2*x)/5 x**2*exp(x)*sin(x)/2 + x**2*exp(x)*cos(x)/2 - x*exp(x)*sin(x) + exp(x)*sin(x)/2 - exp(x)*cos(x)/2 -2*x**(5/2)*cos(x**(1/2)) + 40*x**(3/2)*cos(x**(1/2)) - 240*x**(1/2)*cos(x**(1/2)) + 10*x**2*sin(x**(1/2)) - 120*x*sin(x**(1/2)) + 240*sin(x**(1/2))
4.8は近似値を出せって話なのでscipyを利用。warning出てるけどlimit増やせば誤差は減る。これで大丈夫かは解答も載ってないのでわからず。
import scipy as sp import scipy.integrate import math def f1(x): return math.sin(8. * x) / math.sin(x) print sp.integrate.quad(f1, 0, math.pi / 2.) def f2(x): return (math.exp(2 * x) - 1) / (math.exp(x) + 1) print sp.integrate.quad(f2, 1., 2.) def f3(x): return math.sin(x) ** 2 / x ** 2 print sp.integrate.quad(f3, 0., sp.inf, limit=1000) def f4(x, a): return math.exp(-x) * math.sin(a * x) print sp.integrate.quad(f4, 0., sp.inf, args=(1.)) print sp.integrate.quad(f4, 0., sp.inf, args=(2.))
(1.4476190476190476, 2.3723635847410075e-11) (3.6707742704716053, 4.075378113279188e-14) /Users/tarara/work/py-virtualenvs/numpy-test/lib/python2.7/site-packages/scipy/integrate/quadpack.py:288: UserWarning: The maximum number of subdivisions (1000) has been achieved. If increasing the limit yields no improvement it is advised to analyze the integrand in order to determine the difficulties. If the position of a local difficulty can be determined (singularity, discontinuity) one will probably gain from splitting up the interval and calling the integrator on the subranges. Perhaps a special-purpose integrator should be used. warnings.warn(msg) (1.5707758031670074, 1.0326045566833031e-05) (0.5000000000000002, 1.4875911973447031e-08) (0.39999999999999836, 1.4432856577266494e-08)
