Python Nonlinear Equations with Scipy fsolve

Рет қаралды 63,267

APMonitor.com

8 жыл бұрын

The Scipy optimization package FSOLVE is demonstrated on two introductory problems with 1 and 2 variables.

Пікірлер: 41

@nmana9759 3 жыл бұрын

This is helpful thank you .. Instantly subscribed!

@thaeermsahib3133 3 жыл бұрын

Thank you so much prof. Your lesson very good

@daniel_johnson_pmbpk 2 жыл бұрын

Thanks for sharing! If I have a function which could have anywhere from 1 to 4 solutions, depending on some input parameters, and I want to return all of the solutions, is there a way to do this? I.e. without going through them one at a time, but instead returning a list or array of each of the 1,2,3 or 4 solutions?

@apm 2 жыл бұрын

Sympy returns all solutions but it is only applicable for analytic solutions (simple problems): github.com/APMonitor/data_science/blob/master/10.%20Solve_Equations.ipynb

@yuangao9731 5 жыл бұрын

thank u so much!

@kathytang3006 7 жыл бұрын

Hi Professor, These are great videos, where can i find the videos for other homework?

@apm 7 жыл бұрын

Here are the videos for the other solutions: apmonitor.com/che263/index.php/Main/CourseHomework

@holaram5947 4 жыл бұрын

Thank you sir

@yem.t.3930 3 жыл бұрын

Thanks sir!

@shreeniketjoshi 4 жыл бұрын

Hi, thanks for sharing. I have one question though. @5:23: plt.plot(x, f(x)) plt.plot(x,np.zeros(len(x))) Why did we plot 'x' twice? Or is it something else that is happening? Thanks

@apm 4 жыл бұрын

You always plot x,y data as plt.plot(x,y). It is the values of the horizontal axis. The two trends share the same x values.

@shreeniketjoshi 4 жыл бұрын

@@apm Got it thanks!

@abdulrahmanbres 7 жыл бұрын

Thanks

@matthewerasmus4803 3 жыл бұрын

Dear APMonitor Thank you for uploading these videos. I have learn't a great deal from them. I am trying to find out if you would know how to solve for the nonlinear equations over a range of coefficients. ax^2 + ax + c = 0 where a & c are coefficients. Thus solve for x over the range of a: a = 0:5:1000?

@apm 3 жыл бұрын

For this equation, you could use the quadratic formula but plug in the value of a for b. en.wikipedia.org/wiki/Quadratic_formula

@ruanfreitas4029 6 жыл бұрын

How can i use dependents equations (diffrential equations) using fsolve ?

@apm 6 жыл бұрын

Check out this example problem: apmonitor.com/pdc/index.php/Main/SimulateHIV - see the solution at the end for the source.

@The018fv 6 жыл бұрын

I have a problem with fsolve, sometimes it returns bad solutions I don't know why. I'm solving a function that has a parameter in a loop and it starts out very well, but at some point it returns bad solution eventhough I raise maxfev and lower xtol, what should I do ?

@apm 6 жыл бұрын

+The018fv, it is likely because of poor initial guesses, the problem is very nonlinear and fsolve isn't the right tool, or else the problem doesn't have a solution. I'd recommend that you try to improve the initial guess values (easiest fix) or else use a more capable optimization solver such as shown here apmonitor.com/che263/index.php/Main/PythonOptimization where you can include constraints to help guide the solution.

@frankelindddd 6 жыл бұрын

When I am trying to solve a integral function. It appears: RuntimeWarning: The number of calls to function has reached maxfev. Can I solve it?

@apm 6 жыл бұрын

One easy thing to try is to increase maxfev. See docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.optimize.fsolve.html The option to increase the maximum function evaluations to 500 is: maxfev = 500. There may be another problem with your problem, however, if it needed to reach this limit. I'd recommend simplifying your problem and checking that there is a good initial guess and a solution.

@KazuyaKunXD 3 жыл бұрын

fsolve is very sensitive to initial guess.is there any other way to find root where we can find the root easily?

@apm 3 жыл бұрын

Examples 2 and 3 with the Gekko Optimization Suite show how to use a constrained optimizer for root finding: apmonitor.com/wiki/index.php/Main/GekkoPythonOptimization

@SANAGHUS1934 5 жыл бұрын

Can you kindly show how to plot the graph of the two nonlinear systems?

@apm 5 жыл бұрын

Here is a tutorial that shows how to add multiple trends to a plot: apmonitor.com/che263/index.php/Main/PythonPlots

@17czachA 4 жыл бұрын

What is the error of the fsolve solutions?

@apm 4 жыл бұрын

Could you point to time in the video where you have the question? You can include the reference with something like 5:05 (just include the time).

@Mythias96 6 жыл бұрын

sorry, where is the video for problems 3 and 4??

@apm 6 жыл бұрын

kzfaq.info/get/bejne/bJyYgrCIv8vHYWQ.html (3) and kzfaq.info/get/bejne/j52JdbOirLHWaY0.html (4). All of the assignments and solutions are listed here: apmonitor.com/che263/index.php/Main/CourseHomework This is assignment #14.

@Mythias96 6 жыл бұрын

thank you so much :)

@pythonscienceanddatascienc4351 3 жыл бұрын

Good night, it's me again here from Brazil. I am trying to find two points of intersection between a circle, a line and a parabola, all curves in the plane. There are 3 equations with 2 variables. However, I would like your help to answer a question: - to solve the system I had to put the Z axis in the fsolve. The answer should be zero. Why is it different from zero? - Why, depending on the initial value that I put in fsolve, can he find the point of intersection between 2 curves and not 3? I will send my code. Note that for the initial palpilte (-3, -3, -20)) the answer is [-1.9 -0.9], which is the point of intersection between the circle and the parabola and not between the circle, parabola and the line. import numpy as np from scipy.optimize import fsolve def equations(p): X, Y, Z = p y1 = X-Y**2 +3 # parabolic y2 = X+Y+1 # line y3 = X**2+Y**2-5 # circle return (y1,y2,y3) X, Y, Z = fsolve(equations, (-3, -3, -20)) print(np. around((X,Y),2),Z) # Why the solution is just parabolic and circle? Why Z is not 0? I would appreciate it if you could help me. Luciana

@apm 3 жыл бұрын

The Z value can be anything and result in a successful solution. You may need to switch to an optimizer if you have 2 variables and 3 equations so that you can also impose a condition such as Minimize(Z**2) when it doesn't need to change to match the solution. Here is an updated version with fsolve.

@apm 3 жыл бұрын

import numpy as np from scipy.optimize import fsolve def equations(p): print(p) X, Y, Z = p y1 = X-Y**2 +3 # parabolic y2 = X+Y+1 # line y3 = X**2+Y**2-5 # circle return (y1,y2,y3) # solution 1: -2,-1,Z=anything # solution 2: 1, -2, Z = anything X, Y, Z = fsolve(equations, (-2.5, -1.5, 0)) print(np.around((X,Y),2),Z) # I printed the values p so that you can see what fsolve is guessing

@apm 3 жыл бұрын

Here is an alternative version with gekko (optimizer): from gekko import GEKKO m = GEKKO(remote=False) X,Y,Z = m.Array(m.Var,3) # add constraint to find one solution or the other # for solution 1 Y.lower=0; X.upper=0 m.Equations([X+3==Y**2,X+Y+1==0,X**2+Y**2==5]) m.Minimize(Z**2); m.options.RTOL=0.01 m.solve(disp=False) print(X.value[0],Y.value[0],Z.value[0]) # for solution 2 Y.lower=-1e8; X.upper=1e8 Y.upper=0; X.lower=0 m.solve(disp=False) print(X.value[0],Y.value[0],Z.value[0])

@pythonscienceanddatascienc4351 3 жыл бұрын

@@apm Again, thanks for your quick response and your help. But, I still have the doubt. I took your script and put two initial conditions: (-3, -3, 0) and (-3, -3, -20). The fsolve solution for (-3, -3, -20) does not exist for the 3 equations, only for two. The fsolve solution for the initial guess (-3, -3, 0) already gives a correct answer. I can not understand it. How could fsolve give a solution to only one two equations and not three? It is so confuse for me :(

@rrc 3 жыл бұрын

@@pythonscienceanddatascienc4351 fsolve isn't a very good solver. You just need to give it a better guess value.