Thanks a lot, I have the curse of being a visual learner and this was amazing.
@Carrymejane4 ай бұрын
Woah whole meat
@Carrymejane4 ай бұрын
Thanks!
@Carrymejane4 ай бұрын
This is close to application math, this is so good for gave us the biggest picture of it!
@randolfshemhusain22984 ай бұрын
Visual learning makes things so much better.
@faribasoltani82685 ай бұрын
slope is delta Y divided by delta X
@ImagineCarnage6 ай бұрын
Thank you, this was a great introduction to tge topic.
@plep-m555ww7 ай бұрын
Great video and helpful channel! Khan academy and the organic chemistry guy are getting old and less helpful as school curriculums develop. Super grateful for these simple, direct explanations
@willjadsonevania97878 ай бұрын
teacher I developed a heuristic and would like to share it. My heuristic uses topology and concentric circles. What do you think?.
@ahmedshalaby93438 ай бұрын
more videos please
@ahmedshalaby93438 ай бұрын
in 2 mins just you explained everything
@anityanarayana9 ай бұрын
How to solve this when the x and y are bounded?
@nickokapo976110 ай бұрын
Oh, it's really exciting, i hope to see more videos from you! So far i've seen some of IBM Technology's videos on it, but i'd like to know what the current knowledge and technology of AI is as of september 2023... I wish they'd post longer than 5-8 minute videos on these subjects... It is not that they are not informative, just that they are never in-depth. I remember MIT OpenCourseWare courses were hours upon hours, but nowadays, you do searches about AI, Machine Learning, NLP, Data Analysis, and they're all very general, like an overview at best, it is good for me as a beginner, and i do not promise that i would understand the more extended videos, but i am so curious and intrigued by them, i could easily spend hours learning about it, if only the content was there.
@raminbohlouli196910 ай бұрын
Simple yet extremely informative👍
@fabianb.7429 Жыл бұрын
Just perfect. Thanks
@anshisingh1915 Жыл бұрын
lovely brooo, such good animation, now i have the concept in my head.
@funfair-bs7wf Жыл бұрын
This is a great little video !
@AhmedAhmed-do7im Жыл бұрын
So clear
@marcusaurelius8030 Жыл бұрын
how the hell is this O(n!) ??
@adelsayyahi9665 Жыл бұрын
Thank you, what is the name of the algoodo tolbox you used for simulated annealing?
@MultiJx2 Жыл бұрын
compact and thorough at the same time. thanks !
@yoshitha12 Жыл бұрын
Thank you... ❤
@EXEFaker Жыл бұрын
Imagine being a Salesman and this actually happens (I k it can happen irl on godddd it's a joke)
@kloassie Жыл бұрын
Explain the christofian 1.5 solution and give an heuristic example as well please
@ramyasakthi06 Жыл бұрын
excellent explanation.Thank you so mcuh
@Keyakina Жыл бұрын
But residual != error?
@chinthakawk Жыл бұрын
Works fine in R2021b. % set initial guess values for box dimensions lengthGuess = 1; widthGuess = 1; heightGuess = 1; % load guess values into array x0 = [lengthGuess widthGuess heightGuess]; % call solver to minimize the objective function given the constraint xopt = fmincon(@objective,x0,[],[],[],[],[],[],@constraint,[]) % retrieve optimized box sizing and volume volumeOpt = calcVolume(xopt) % calculate surface area with optimized values just to double check surfaceAreaOpt = calcSurface(xopt) % define function to calculate volume of box function volume = calcVolume(x) length = x(1); width = x(2); height = x(3); volume = length * width * height; end % define function to calculate surface area of box function surfaceArea = calcSurface(x) length = x(1); width = x(2); height = x(3); surfaceArea = 2*length*width + 2*length*height + 2*height*width; end % define objective function for optimization function obj = objective(x) obj = -calcVolume(x); end % define constraint for optimization function [c, ceq] = constraint(x) c = calcSurface(x) - 10; ceq = []; end
@ernstuzhansky Жыл бұрын
Excellent! Thank you.
@tomerweinbach4059 Жыл бұрын
great explanation!
@SwanPrncss Жыл бұрын
Omg, your explanation is better than other youtube videos and my teacher because I'm a visual learner.
@DiditCoding Жыл бұрын
Noice
@alice20001 Жыл бұрын
Why use the squares instead of the absolute values?
@laraelnourr Жыл бұрын
because they are easier to compute and deal with mathematically. But we can use absolute values too!
@faheemrasheed9967 Жыл бұрын
because it gives more clear picture if we have error of ,1 and if we square it it will give 0,01 which is kind of scaled.
@AchiragChiragg6 ай бұрын
@@faheemrasheed9967actually it's the other way around. It's better to use absolute value instead of squares as it can amplify the outliers and influence the final fit.
@andrea-mj9ce Жыл бұрын
The Nelder-Mead method is not explained long enough to understand it.
@VictoriaOtunsha Жыл бұрын
Thank you for simplifying this
@jedediah-fanuel Жыл бұрын
<3
@jedediah-fanuel Жыл бұрын
<3
@rudypieplenbosch6752 Жыл бұрын
no example, pretty useless
@tankkinnari Жыл бұрын
Good explanation 👍
@tonmoysharma5758 Жыл бұрын
Excellent video and also quite easy to understand
@markneuhold7065 Жыл бұрын
This is a very similar problem I have but I have a linear obj func and constraint, and constraint is an equality. When using SLSQP I get error "singular matrix c in lsq subproblem". Seems I should use linprog but I'm not sure how or whether this type of problem can be converted to linprog. Any ideas?
@josephdorman8010 Жыл бұрын
I suppose we can also say optimization is choosing the best input or best process, or both the best process and best input to yield the best output
@thatgameguy4929 Жыл бұрын
Have you tried slime mold?
@sitrakaforler8696 Жыл бұрын
Nice 👍🏽
@lugaseth3732 Жыл бұрын
Thank you for great videos. Concise, engaging, and clear explanations.
@Jkauppa2 жыл бұрын
newtons method is gradient descent
@Jkauppa2 жыл бұрын
many randomized starts to solve most of the issues, ie, grid starting location search
@Jkauppa2 жыл бұрын
if you have a higher value somewhere and lower somewhere, you are guaranteed to have the middle values between through some route (1d or 2d, or multi dimensional)
@Jkauppa2 жыл бұрын
so you dont get stuck in the flat derivatives, you can use the point-to-point derivatives, roughly, to get a next useful point
@Jkauppa2 жыл бұрын
the grid search with some binary search narrowing criteria, without knowing anything about gradients or derivatives
@Jkauppa2 жыл бұрын
if you have gradient or derivative of zero, then you either have fully horizontal plane or line, or are guaranteed to have either lower or higher values left and right
@Jkauppa2 жыл бұрын
fit a function f(x) to data with normal noise, f(x) can be a line, or a polynomial, etc, includes outlier handling, least squares is very sus
@Jkauppa2 жыл бұрын
line with normal noise is a better answer than just a line
@Jkauppa2 жыл бұрын
constant std additive normal noise assumed
@Jkauppa2 жыл бұрын
think like you are removing a base line function from the data points, either linear (like pca, f(x)=kx+c) or polynome (nonlinear pca, f(x)=...+ax^2+bx+c), then checking if the noise is from a normal distribution, ie, trying to make the noise after removing the base line as normal as possible, if you do linear, the noise might not be normal, so you get only a partial pca component fit, kinda
@neerajmahapatra52392 жыл бұрын
This is amazing video! Lots of you tubers just start teaching the logic and solving the problem without stating the usage of the problem by relating it to real life scenarios.
@Jkauppa2 жыл бұрын
try sorting all edge lengths, amount ½n^2, n is location count, then try the permutations until you have a guaranteed shortest loop path
@Jkauppa2 жыл бұрын
so you travel edges, not permutating the target (cities, locations)
@Jkauppa2 жыл бұрын
it gives even more permutations to test, but gives an actual solution
@Jkauppa2 жыл бұрын
graph theory solution
@Jkauppa2 жыл бұрын
try djisktra shortest path algorithm, breadth first on all location starting points
@Jkauppa2 жыл бұрын
please note, not all permutations are unique routes
Пікірлер
Superb!
drink some water man
u can say thanks instead of this
clear and brief idea
Thanks a lot, I have the curse of being a visual learner and this was amazing.
Woah whole meat
Thanks!
This is close to application math, this is so good for gave us the biggest picture of it!
Visual learning makes things so much better.
slope is delta Y divided by delta X
Thank you, this was a great introduction to tge topic.
Great video and helpful channel! Khan academy and the organic chemistry guy are getting old and less helpful as school curriculums develop. Super grateful for these simple, direct explanations
teacher I developed a heuristic and would like to share it. My heuristic uses topology and concentric circles. What do you think?.
more videos please
in 2 mins just you explained everything
How to solve this when the x and y are bounded?
Oh, it's really exciting, i hope to see more videos from you! So far i've seen some of IBM Technology's videos on it, but i'd like to know what the current knowledge and technology of AI is as of september 2023... I wish they'd post longer than 5-8 minute videos on these subjects... It is not that they are not informative, just that they are never in-depth. I remember MIT OpenCourseWare courses were hours upon hours, but nowadays, you do searches about AI, Machine Learning, NLP, Data Analysis, and they're all very general, like an overview at best, it is good for me as a beginner, and i do not promise that i would understand the more extended videos, but i am so curious and intrigued by them, i could easily spend hours learning about it, if only the content was there.
Simple yet extremely informative👍
Just perfect. Thanks
lovely brooo, such good animation, now i have the concept in my head.
This is a great little video !
So clear
how the hell is this O(n!) ??
Thank you, what is the name of the algoodo tolbox you used for simulated annealing?
compact and thorough at the same time. thanks !
Thank you... ❤
Imagine being a Salesman and this actually happens (I k it can happen irl on godddd it's a joke)
Explain the christofian 1.5 solution and give an heuristic example as well please
excellent explanation.Thank you so mcuh
But residual != error?
Works fine in R2021b. % set initial guess values for box dimensions lengthGuess = 1; widthGuess = 1; heightGuess = 1; % load guess values into array x0 = [lengthGuess widthGuess heightGuess]; % call solver to minimize the objective function given the constraint xopt = fmincon(@objective,x0,[],[],[],[],[],[],@constraint,[]) % retrieve optimized box sizing and volume volumeOpt = calcVolume(xopt) % calculate surface area with optimized values just to double check surfaceAreaOpt = calcSurface(xopt) % define function to calculate volume of box function volume = calcVolume(x) length = x(1); width = x(2); height = x(3); volume = length * width * height; end % define function to calculate surface area of box function surfaceArea = calcSurface(x) length = x(1); width = x(2); height = x(3); surfaceArea = 2*length*width + 2*length*height + 2*height*width; end % define objective function for optimization function obj = objective(x) obj = -calcVolume(x); end % define constraint for optimization function [c, ceq] = constraint(x) c = calcSurface(x) - 10; ceq = []; end
Excellent! Thank you.
great explanation!
Omg, your explanation is better than other youtube videos and my teacher because I'm a visual learner.
Noice
Why use the squares instead of the absolute values?
because they are easier to compute and deal with mathematically. But we can use absolute values too!
because it gives more clear picture if we have error of ,1 and if we square it it will give 0,01 which is kind of scaled.
@@faheemrasheed9967actually it's the other way around. It's better to use absolute value instead of squares as it can amplify the outliers and influence the final fit.
The Nelder-Mead method is not explained long enough to understand it.
Thank you for simplifying this
<3
<3
no example, pretty useless
Good explanation 👍
Excellent video and also quite easy to understand
This is a very similar problem I have but I have a linear obj func and constraint, and constraint is an equality. When using SLSQP I get error "singular matrix c in lsq subproblem". Seems I should use linprog but I'm not sure how or whether this type of problem can be converted to linprog. Any ideas?
I suppose we can also say optimization is choosing the best input or best process, or both the best process and best input to yield the best output
Have you tried slime mold?
Nice 👍🏽
Thank you for great videos. Concise, engaging, and clear explanations.
newtons method is gradient descent
many randomized starts to solve most of the issues, ie, grid starting location search
if you have a higher value somewhere and lower somewhere, you are guaranteed to have the middle values between through some route (1d or 2d, or multi dimensional)
so you dont get stuck in the flat derivatives, you can use the point-to-point derivatives, roughly, to get a next useful point
the grid search with some binary search narrowing criteria, without knowing anything about gradients or derivatives
if you have gradient or derivative of zero, then you either have fully horizontal plane or line, or are guaranteed to have either lower or higher values left and right
fit a function f(x) to data with normal noise, f(x) can be a line, or a polynomial, etc, includes outlier handling, least squares is very sus
line with normal noise is a better answer than just a line
constant std additive normal noise assumed
think like you are removing a base line function from the data points, either linear (like pca, f(x)=kx+c) or polynome (nonlinear pca, f(x)=...+ax^2+bx+c), then checking if the noise is from a normal distribution, ie, trying to make the noise after removing the base line as normal as possible, if you do linear, the noise might not be normal, so you get only a partial pca component fit, kinda
This is amazing video! Lots of you tubers just start teaching the logic and solving the problem without stating the usage of the problem by relating it to real life scenarios.
try sorting all edge lengths, amount ½n^2, n is location count, then try the permutations until you have a guaranteed shortest loop path
so you travel edges, not permutating the target (cities, locations)
it gives even more permutations to test, but gives an actual solution
graph theory solution
try djisktra shortest path algorithm, breadth first on all location starting points
please note, not all permutations are unique routes