I have a set of measured radii (t+epsilon+error) at an equally spaced angles.
The model is circle of radius (R) with center at (r, Alpha) with added small noise and some random error values which are much bigger than noise.
The problem is to find the center of the circle model (r,Alpha) and the radius of the circle (R). But it should not be too much sensitive to random error (in below data points at 7 and 14).
Some radii could be missing therefore the simple mean would not work here.
I tried least square optimization but it significantly reacts on error.
Is there a way to optimize least deltas but not the least squares of delta in Python?
Model:
n=36
R=100
r=10
Alpha=2*Pi/6
Data points:
[95.85, 92.66, 94.14, 90.56, 88.08, 87.63, 88.12, 152.92, 90.75, 90.73, 93.93, 92.66, 92.67, 97.24, 65.40, 97.67, 103.66, 104.43, 105.25, 106.17, 105.01, 108.52, 109.33, 108.17, 107.10, 106.93, 111.25, 109.99, 107.23, 107.18, 108.30, 101.81, 99.47, 97.97, 96.05, 95.29]
It seems like your main problem here is going to be removing outliers. There are a couple of ways to do this, but for your application, your best bet is to probably just to remove items based on their distance from the median (Since the median is much less sensitive to outliers than the mean.)
If you're using numpy that would looks like this:
def remove_outliers(data_points, margin=1.5):
nd = np.abs(data_points - np.median(data_points))
s = nd/np.median(nd)
return data_points[s<margin]
After which you should run least squares.
If you're not using numpy you can do something similar with native python lists:
def median(points):
return sorted(points)[len(points)/2] # evaluates to an int in python2
def remove_outliers(data_points, margin=1.5):
m = median(data_points)
centered_points = [abs(point - m) for point in data_points]
centered_median = median(centered_points)
ratios = [datum/centered_median for datum in centered_points]
return [point for i, point in enumerate(data_points) if ratios[i]>margin]
If you're looking to just not count outliers as highly you can just calculate the mean of your dataset, which is just a linear equivalent of the least-squares optimization.
If you're looking for something a little better I might suggest throwing your data through some kind of low pass filter, but I don't think that's really needed here.
A low-pass filter would probably be the best, which you can do as follows: (Note, alpha is a number you will have to fiddle with to get your desired output.)
def low_pass(data, alpha):
new_data = [data[0]]
for i in range(1, len(data)):
new_data.append(alpha * data[i] + (1 - alpha) * new_data[i-1])
return new_data
At which point your least squares optimization should work fine.
Replying to your final question
Is there a way to optimize least deltas but not the least squares of delta in Python?
Yes, pick an optimization method (for example downhill simplex implemented in scipy.optimize.fmin) and use the sum of absolute deviations as a merit function. Your dataset is small, I suppose that any general purpose optimization method will converge quickly. (In case of non-linear least squares fitting it is also possible to use general purpose optimization algorithm, but it's more common to use the Levenberg-Marquardt algorithm which minimizes sums of squares.)
If you are interested when minimizing absolute deviations instead of squares has theoretical justification see Numerical Recipes, chapter Robust Estimation.
From practical side, the sum of absolute deviations may not have unique minimum.
In the trivial case of two points, say, (0,5) and (1,9) and constant function y=a, any value of a between 5 and 9 gives the same sum (4). There is no such problem when deviations are squared.
If minimizing absolute deviations would not work, you may consider heuristic procedure to identify and remove outliers. Such as RANSAC or ROUT.
Related
I am estimating the fundamental matrix and the essential matrix by using the inbuilt functions in opencv.I provide input points to the function by using ORB and brute force matcher.These are the problems that i am facing:
1.The essential matrix that i compute from in built function does not match with the one i find from mathematical computation using fundamental matrix as E=k.t()FK.
2.As i vary the number of points used to compute F and E,the values of F and E are constantly changing.The function uses Ransac method.How do i know which value is the correct one??
3.I am also using an inbuilt function to decompose E and find the correct R and T from the 4 possible solutions.The value of R and T also change with the changing E.More concerning is the fact that the direction vector T changes without a pattern.Say it was in X direction at a value of E,if i change the value of E ,it changes to Y or Z.Y is this happening????.Has anyone else had the same problem.???
How do i resolve this problem.My project involves taking measurements of objects from images.
Any suggestions or help would be welcome!!
Both F and E are defined up to a scale factor. It may help to normalize the matrices, e. g. by dividing by the last element.
RANSAC is a randomized algorithm, so you will get a different result every time. You can test how much it varies by triangulating the points, or by computing the reprojection errors. If the results vary too much, you may want to increase the number of RANSAC trials or decrease the distance threshold, to make sure that RANSAC converges to the correct solution.
Yes, Computing Fundamental Matrix gives a different matrix every time as it is defined up to a scale factor.
It is a Rank 2 matrix with 7DOF(3 rot, 3 trans, 1 scaling).
The fundamental matrix is a 3X3 matrix, F33(3rd col and 3rd row) is scale factor.
You make ask why do we append matrix with constant at F33, Because of (X-Left)F(x-Right)=0, This is a homogenous equation with infinite solutions, we are adding a constraint by making F33 constant.
Given a 2D point p, I'm trying to calculate the smallest distance between that point and a functional curve, i.e., find the point on the curve which gives me the smallest distance to p, and then calculate that distance. The example function that I'm using is
f(x) = 2*sin(x)
My distance function for the distance between some point p and a provided function is
def dist(p, x, func):
x = np.append(x, func(x))
return sum([[i - j]**2 for i,j in zip(x,p)])
It takes as input, the point p, a position x on the function, and the function handle func. Note this is a squared Euclidean distance (since minimizing in Euclidean space is the same as minimizing in squared Euclidean space).
The crucial part of this is that I want to be able to provide bounds for my function so really I'm finding the closest distance to a function segment. For this example my bounds are
bounds = [0, 2*np.pi]
I'm using the scipy.optimize.minimize function to minimize my distance function, using the bounds. A result of the above process is shown in the graph below.
This is a contour plot showing distance from the sin function. Notice how there appears to be a discontinuity in the contours. For convenience, I've plotted a few points around that discontinuity and the "closet" points on the curve that they map to.
What's actually happening here is that the scipy function is finding a local minimum (given some initial guess), but not a global one and that is causing the discontinuity. I know finding the global minimum of any function is impossible, but I'm looking for a more reliable way to find the global minimum.
Possible methods for finding a global minimum would be
Choose a smart initial guess, but this amounts to knowing approximately where the global minimum is to begin with, which is using the solution of the problem to solve it.
Use a multiple initial guesses and choose the answer which gets to the best minimum. This however seems like a poor choice, especially when my functions get more complicated (and higher dimensional).
Find the minimum, then perturb the solution and find the minimum again, hoping that I may have knocked it into a better minimum. I'm hoping that maybe there is some way to do this simply without evoking some complicated MCMC algorithm or something like that. Speed counts for this process.
Any suggestions about the best way to go about this, or possibly directions to useful functions that may tackle this problem would be great!
As suggest in a comment, you could try a global optimization algorithm such as scipy.optimize.differential_evolution. However, in this case, where you have a well-defined and analytically tractable objective function, you could employ a semi-analytical approach, taking advantage of the first-order necessary conditions for a minimum.
In the following, the first function is the distance metric and the second function is (the numerator of) its derivative w.r.t. x, that should be zero if a minimum occurs at some 0<x<2*np.pi.
import numpy as np
def d(x, p):
return np.sum((p-np.array([x,2*np.sin(x)]))**2)
def diff_d(x, p):
return -2 * p[0] + 2 * x - 4 * p[1] * np.cos(x) + 4 * np.sin(2*x)
Now, given a point p, the only potential minimizers of d(x,p) are the roots of diff_d(x,p) (if any), as well as the boundary points x=0 and x=2*np.pi. It turns out that diff_d may have more than one root. Noting that the derivative is a continuous function, the pychebfun library offers a very efficient method for finding all the roots, avoiding cumbersome approaches based on the scipy root-finding algorithms.
The following function provides the minimum of d(x, p) for a given point p:
import pychebfun
def min_dist(p):
f_cheb = pychebfun.Chebfun.from_function(lambda x: diff_d(x, p), domain = (0,2*np.pi))
potential_minimizers = np.r_[0, f_cheb.roots(), 2*np.pi]
return np.min([d(x, p) for x in potential_minimizers])
Here is the result:
I'm trying to do a K-means clustering of some dataset using sklearn. The problem is that one of the dimensions is hour-of-day: a number from 0-23 and so the distance algorithm then thinks that 0 is very far from 23, because in absolute terms it is. In reality and for my purposes, hour 0 is very close to hour 23. Is there a way to make the distance algorithm do some form of wrap-around so it computes the more 'real' time difference.
I'm doing something simple, similar to the following:
from sklearn.cluster import KMeans
clusters = KMeans(n_clusters = 2)
data = vstack(data)
fit = clusters.fit(data)
classes = fit.predict(data)
data elements looks something like [22, 418, 192] where the first element is the hour.
Any ideas?
Even though #elyase answer is accepted, I think it is not the correct approach.
Yes, to use such distance you have to refine your distance measure and so - use different library. But what is more important - concept of mean used in k-means won't suit the cyclic dimension. Lets consider following example:
#current cluster X,, based on centroid position Xc=24
x1=1
x2=24
#current cluster Y, based on centroid position Yc=10
y1=12
y2=13
computing simple arithmetic mean will place the centoids in Xc=12.5,Yc=12.5, which from the point of view of cyclic meausre is incorect, it should be Xc=0.5,Yc=12.5. As you can see, asignment based on the cyclic distance measure is not "compatible" with simple mean operation, and leads to bizzare results.
Simple k-means will result in clusters {x1,y1}, {x2,y2}
Simple k--means + distance measure result in degenerated super cluster {x1,x2,y1,y2}
Correct clustering would be {x1,x2},{y1,y2}
Solving this problem requires checking one if (whether it is better to measure "simple average" or by representing one of the points as x'=x-24). Unfortunately given n points it makes 2^n possibilities.
This seems as a use case of the kernelized k-means, where you are actually clustering in the abstract feature space (in your case - a "tube" rolled around the time dimension) induced by kernel ("similarity measure", being the inner product of some vector space).
Details of the kernel k-means are given here
Why k-means doesn't work with arbitrary distances
K-means is not a distance-based algorithm.
K-means minimizes the Within-Cluster-Sum-of-Squares, which is a kind of variance (it's roughly the weighted average variance of all clusters, where each object and dimension is given the same weight).
In order for Lloyds algorithm to converge you need to have both steps optimize the same function:
the reassignment step
the centroid update step
Now the "mean" function is a least-squares estimator. I.e. choosing the mean in step 2 is optimal for the WCSS objective. Assigning objects by least-squares deviation (= squared Euclidean distance, monotone to Euclidean distance) in step 1 also yields guaranteed convergence. The mean is exactly where your wrap-around idea would fall apart.
If you plug in a random other distance function as suggested by #elyase k-means might no longer converge.
Proper solutions
There are various solutions to this:
Use K-medoids (PAM). By choosing the medoid instead of the mean you do get guaranteed convergence with arbitrary distances. However, computing the medoid is rather expensive.
Transform the data into a kernel space where you are happy with minimizing Sum-of-Squares. For example, you could transform the hour into sin(hour / 12 * pi), cos(hour / 12 * pi) which may be okay for SSQ.
Use other, distance-based clustering algorithms. K-means is old, and there has been a lot of research on clustering since. You may want to start with hierarchical clustering (which actually is just as old as k-means), and then try DBSCAN and the variants of it.
The easiest approach, to me, is to adapt the K-means algorithm wraparound dimension via computing the "circular mean" for the dimension. Of course, you will also need to change the distance-to-centroid calculation accordingly.
#compute the mean of hour 0 and 23
import numpy as np
hours = np.array(range(24))
#hours to angles
angles = hours/24 * (2*np.pi)
sin = np.sin(angles)
cos = np.cos(angles)
a = np.arctan2(sin[23]+sin[0], cos[23]+cos[0])
if a < 0: a += 2*np.pi
#angle back to hour
hour = a * 24 / (2*np.pi)
#23.5
I'd like to linearly fit the data that were NOT sampled independently. I came across generalized least square method:
b=(X'*V^(-1)*X)^(-1)*X'*V^(-1)*Y
The equation is Matlab format; X and Y are coordinates of the data points, and V is a "variance matrix".
The problem is that due to its size (1000 rows and columns), the V matrix becomes singular, thus un-invertable. Any suggestions for how to get around this problem? Maybe using a way of solving generalized linear regression problem other than GLS? The tools that I have available and am (slightly) familiar with are Numpy/Scipy, R, and Matlab.
Instead of:
b=(X'*V^(-1)*X)^(-1)*X'*V^(-1)*Y
Use
b= (X'/V *X)\X'/V*Y
That is, replace all instances of X*(Y^-1) with X/Y. Matlab will skip calculating the inverse (which is hard, and error prone) and compute the divide directly.
Edit: Even with the best matrix manipulation, some operations are not possible (for example leading to errors like you describe).
An example of that which may be relevant to your problem is if try to solve least squares problem under the constraint the multiple measurements are perfectly, 100% correlated. Except in rare, degenerate cases this cannot be accomplished, either in math or physically. You need some independence in the measurements to account for measurement noise or modeling errors. For example, if you have two measurements, each with a variance of 1, and perfectly correlated, then your V matrix would look like this:
V = [1 1; ...
1 1];
And you would never be able to fit to the data. (This generally means you need to reformulate your basis functions, but that's a longer essay.)
However, if you adjust your measurement variance to allow for some small amount of independence between the measurements, then it would work without a problem. For example, 95% correlated measurements would look like this
V = [1 0.95; ...
0.95 1 ];
You can use singular value decomposition as your solver. It'll do the best that can be done.
I usually think about least squares another way. You can read my thoughts here:
http://www.scribd.com/doc/21983425/Least-Squares-Fit
See if that works better for you.
I don't understand how the size is an issue. If you have N (x, y) pairs you still only have to solve for (M+1) coefficients in an M-order polynomial:
y = a0 + a1*x + a2*x^2 + ... + am*x^m
Good Morning,
I am implimenting a Cressman filter for doing distance weighted averages in Numpy.. I use a Ball Tree implimentation (thanks to Jake VanderPlas) to return a list of locatations for each point in a request array.. the query array (q) is shape [n,3] and at each point has the x,y,z at point I want to do a weighted average of points stored in the tree.. the code wrapped around the tree returns points within a certain distance so I get an arrays of variable length arrays..
I use a where to find non-empty entries (ie positions where there were at least some points within the radius of influence) creating the isgood array...
I then loop over all query points to return the weighted average of the values self.z (note that this can either be dims=1 or dims=2 to allow multiple co-gridding)
so the thing that complilcates using map or other quicker methods is the nonuniformity of the lengths of the arrays within self.distances and self.locations... I am still fairly green to numpy/python but I can not think of a way to do this array wise (ie not reverting to loops)
self.locations, self.distances = self.tree.query_radius( q, r, return_distance=True)
t2=time()
if debug: print "Removing voids"
isgood=np.where( np.array([len(x) for x in self.locations])!=0)[0]
interpol = np.zeros( (len(self.locations),) + np.shape(self.z[0]) )
interpol.fill(np.nan)
for dist, ix, posn, roi in zip(self.distances[isgood], self.locations[isgood], isgood, r[isgood]):
interpol[isgood[jinterpol]] = np.average(self.z[ix], weights=(roi**2-dist**2) / (roi**2 + dist**2), axis=0)
jinterpol += 1
so... Any hints of how to speed up the loop?..
For a typical mapping as appied to mapping weather radar data from a range,azimuth,elevation grid to a cartesian grid where I have 240x240x34 points and 4 variables takes 99s to query the tree (written by Jake in C and cython.. this is the hard step as you need to search the data!) and 100 seconds to do the calculation... which in my opinon is slow?? where is my overhead? is np.mean efficient or as it is called millions of times is there a speedup to be gained here? would I gain by using float32 rather than the default64... or even scaling to ints (which would be very hard to avoid wrap around in the weighting... any hints gratefully recieved!
You can find a discussion about the relative merits of the Cressman scheme vs using a Gaussian weight function at:
http://www.flame.org/~cdoswell/publications/radar_oa_00.pdf
The key is to match the smoothing parameter to the data (I recommend using a value close to the average spacing between data points). Once you know the smoothing parameter, you can set an "influence radius" equal to the radius where the weight function falls to 0.01 (or whatever).
How important is speed? If you wish, rather than calling an exponential function to determine the weight, you can make up a discrete table of weights for some fixed number of radius increments, which speeds up the calculation considerably. Ideally, you should have data outside the grid boundaries that can be used in the mapping of the values surrounding the gridpoints (even on the boundary points of the grid). Note this is NOT a true interpolation scheme - it won't return the observed values at the data points exactly. Like the Cressman scheme, it's a low-pass filer.