Yesterday, I posted a question about general concept of SVM Primal Form Implementation:
Support Vector Machine Primal Form Implementation
and "lejlot" helped me out to understand that what I am solving is a QP problem.
But I still don't understand how my objective function can be expressed as QP problem
(http://en.wikipedia.org/wiki/Support_vector_machine#Primal_form)
Also I don't understand how QP and Quasi-Newton method are related
All I know is Quasi-Newton method will SOLVE my QP problem which supposedly formulated from
my objective function (which I don't see the connection)
Can anyone walk me through this please??
For SVM's, the goal is to find a classifier. This problem can be expressed in terms of a function that you are trying to minimize.
Let's first consider the Newton iteration. Newton iteration is a numerical method to find a solution to a problem of the form f(x) = 0.
Instead of solving it analytically we can solve it numerically by the follwing iteration:
x^k+1 = x^k - DF(x)^-1 * F(x)
Here x^k+1 is the k+1th iterate, DF(x)^-1 is the inverse of the Jacobian of F(x) and x is the kth x in the iteration.
This update runs as long as we make progress in terms of step size (delta x) or if our function value approaches 0 to a good degree. The termination criteria can be chosen accordingly.
Now consider solving the problem f'(x)=0. If we formulate the Newton iteration for that, we get
x^k+1 = x - HF(x)^-1 * DF(x)
Where HF(x)^-1 is the inverse of the Hessian matrix and DF(x) the gradient of the function F. Note that we are talking about n-dimensional Analysis and can not just take the quotient. We have to take the inverse of the matrix.
Now we are facing some problems: In each step, we have to calculate the Hessian matrix for the updated x, which is very inefficient. We also have to solve a system of linear equations, namely y = HF(x)^-1 * DF(x) or HF(x)*y = DF(x).
So instead of computing the Hessian in every iteration, we start off with an initial guess of the Hessian (maybe the identity matrix) and perform rank one updates after each iterate. For the exact formulas have a look here.
So how does this link to SVM's?
When you look at the function you are trying to minimize, you can formulate a primal problem, which you can the reformulate as a Dual Lagrangian problem which is convex and can be solved numerically. It is all well documented in the article so I will not try to express the formulas in a less good quality.
But the idea is the following: If you have a dual problem, you can solve it numerically. There are multiple solvers available. In the link you posted, they recommend coordinate descent, which solves the optimization problem for one coordinate at a time. Or you can use subgradient descent. Another method is to use L-BFGS. It is really well explained in this paper.
Another popular algorithm for solving problems like that is ADMM (alternating direction method of multipliers). In order to use ADMM you would have to reformulate the given problem into an equal problem that would give the same solution, but has the correct format for ADMM. For that I suggest reading Boyds script on ADMM.
In general: First, understand the function you are trying to minimize and then choose the numerical method that is most suited. In this case, subgradient descent and coordinate descent are most suited, as stated in the Wikipedia link.
Related
I have a two 3D variables for a each time step (so I have N 3d matrix var(Nx,Ny,Nz), for each variables). I want to construct the two point statistics but I guess I'm doing something wrong.
Two-point statistics formula, where x_r is the reference point and x is the independent variable
I know that the theoretical formulation of a two-point cross correlation is the one written above.
Let's for sake of simplicity ignore the normalization, so I'm focusing on the numerator, that is the part I'm struggling with.
So, my two variables are two 3D matrix, with the following notation phi(x,y,z) = phi(i,j,k), same for psi.
My aim is to compute a 3d correlation, so given a certain reference point Reference_Point = (xr,yr,zr), but I guess I'm doing something wrong. I'm trying that on MATLAB, but my results are not accurate, and by doing some researches online it does seem that I should do convolutions or fft, but I don't find any theoretical framework that explains how to do that and why the formulation above in practices should be implemented by the use of a conv or fft. Moreover I would like to implement my cross-correlation in the spatial domain and not in the frequency domain, and with the convolution I don't understand how to choose the reference point.
Thank you so much in advance for reply
I've to fit the following exponential function to a time-series data (data).
$C(t)$ = $C_{\infty} (1-\exp(-\frac{t}{\tau}))$
I want to compute the time scale $\tau$ at which C(t) reaches $C_{\infty}$. I would like to ask for suggestions on how $\tau$ can be computed. I found an example here that use curve fitting. But I am not sure how to use curve_fit library in scipy to set up the problem described above.
One cannot expect a good fitting along the whole curve with the function that you choose.
This is because especially at t=0 this function returns C=0 while the data value is C=2.5 .This is very far considering the order of magnitude.
Nevertheless on can try to fit this function for a rough result. A non-linear regression calculus is necessary : this is the usual approach using available softwares. This is the recommended method in context of academic exercices.
Alternatively and more simply, a linear regression can be used thanks to a non-conventional method explained in https://fr.scribd.com/doc/14674814/Regressions-et-equations-integrales .
The result is shown below.
For a better fitting one have to take account of the almost constant value of data in the neighborhood of t=0. Choosing a function made of two logistic functions would be recommended. But the calculus is more complicated.
IN ADDITION, AFTER THE OP CHANGES THE DATA :
The change of data makes out of date the above answer.
In fact artificially changing the origin of the y-scale so that y=0 at t=0 changes nothing. The slope at t=0 of the chosen fonction is far to be nul, while the slope of the data curve is almost 0. This remains incompatible.
Definitively the chosen function y=C*(1-exp(-t/tau)) cannot fit correctly the data (the preceeding data or the new data as well).
As already pointed out, for a better fitting one have to take account of the almost constant value of data in the neighborhood of t=0. Choosing a function made of two logistic functions would be recommended. But the calculus is more complicated.
I was wondering if there was a direct way of computing the iteration matrix for nth Linear Block Gauss Seidel iteration within OpenMDAO?
thank you
If I understand you correctly, you are referring to the matrix-form of the Gauss Seidel algorithm where you take Ax=b, and break A up into the Diagonal (D), Lower (L) and Upper (U) parts, then use those parts to compute the next iterate.
Specifically you compute [D-L]^-1. This, I believe is what you are referring to as the "iteration matrix" (I am not familiar with this terminology, but based on the algorithm I'm comfortable making an educated guess).
This formulation of the algorithm is useful to think about and a simple way to implement it, but OpenMDAO takes a different approach. The LBGS algorithm implemented in OpenMDAO is set up to work in a matrix-free manner. That means it only interacts with the linear operator methods solve_linear and apply_linear and never explicitly assembles the A matrix at all. Hence there isn't an opportunity to split A up into D, L, U.
Depending on the way you constructed the model, the A matrix you would need might or might not be there at all because OpenMDAO is capable of working in a completely matrix free context. However, if all of your components use the compute_partials or linearize methods to provide partial derivatives then the data you would need for the A matrix does exist in memory.
You'll have to dig for it a bit, and ironically the best place to see how to do that is in the direct solver which does actually require the matrix be formed to compute a factorization.
Also, in that code you'll see a function can iteratively call the linear operator to construct a dense matrix even if the underlying components don't provide their partials directly. Please note that this approach for assembling the matrix is extremely slow and is not recommended for normal operations.
Let's say, I have two random variables,x and y, both of them have n observations. I've used a forecasting method to estimate xn+1 and yn+1, and I also got the standard error for both xn+1 and yn+1. So my question is that what the formula would be if I want to know the standard error of xn+1 + yn+1, xn+1 - yn+1, (xn+1)*(yn+1) and (xn+1)/(yn+1), so that I can calculate the prediction interval for the 4 combinations. Any thought would be much appreciated. Thanks.
Well, the general topic you need to look at is called "change of variables" in mathematical statistics.
The density function for a sum of random variables is the convolution of the individual densities (but only if the variables are independent). Likewise for the difference. In special cases, that convolution is easy to find. For example, for Gaussian variables the density of the sum is also a Gaussian.
For product and quotient, there aren't any simple results, except in special cases. For those, you might as well compute the result directly, maybe by sampling or other numerical methods.
If your variables x and y are not independent, that complicates the situation. But even then, I think sampling is straightforward.
here is what I want to do (preferably with Matlab):
Basically I have several traces of cars driving on an intersection. Each one is noisy, so I want to take the mean over all measurements to get a better approximation of the real route. In other words, I am looking for a way to approximate the Curve, which has the smallest distence to all of the meassured traces (in a least-square sense).
At the first glance, this is quite similar what can be achieved with spap2 of the CurveFitting Toolbox (good example in section Least-Squares Approximation here).
But this algorithm has some major drawback: It assumes a function (with exactly one y(x) for every x), but what I want is a curve in 2d (which may have several y(x) for one x). This leads to problems when cars turn right or left with more then 90 degrees.
Futhermore it takes the vertical offsets and not the perpendicular offsets (according to the definition on wolfram).
Has anybody an idea how to solve this problem? I thought of using a B-Spline and change the number of knots and the degree until I reached a certain fitting quality, but I can't find a way to solve this problem analytically or with the functions provided by the CurveFitting Toolbox. Is there a way to solve this without numerical optimization?
mbeckish is right. In order to get sufficient flexibility in the curve shape, you must use a parametric curve representation (x(t), y(t)) instead of an explicit representation y(x). See Parametric equation.
Given n successive points on the curve, assign them their true time if you know it or just integers 0..n-1 if you don't. Then call spap2 twice with vectors T, X and T, Y instead of X, Y. Now for arbitrary t you get a point (x, y) on the curve.
This won't give you a true least squares solution, but should be good enough for your needs.