For donkeys. No parking

For donkeys. No parking

For donkeys. No parking.

This is the gate of a small computer shop in Almaty, Kazakhstan.

Advertisements
Published in: on 29/08/2015 at 18:29  Leave a Comment  

Matlab — Scilab survival phrasebook

This is a one-page dictionary in case you need to switch between Matlab and Scilab.

Matlab Scilab Function
acot(A) atan((1) ./A) Inverse cotangent
acoth(A) atanh((1) ./A) Inverse hyperbolic cotangent
acsc(A) asin((1) ./A) Inverse cosecant
acsch(A) asinh((1) ./A) Inverse hyperbolic cosecant
all and Test to determine if all elements are nonzero
angle(A) atan(imag(A),real(A)) Phase angle
any or Test to determine if any nonzeros elements
asec(A) acos((1) ./A) Inverse secant
asech(A) acosh((1) ./A) Inverse hyperbolic secant
atan2(y,x) atan(y,x) Four-quadrant inverse tangent
[T,Ab]=balance(A) [Ab,T]=balanc(A) Diagonal scaling to improve eigenvalue accuracy
blkdiag sysdiag Construct block diagonal matrix from input arguments
cot cotg Cotangent
cputime timer() Elapsed CPU time
csc(A) (1) ./sin(A) Cosecant
csch(A) (1) ./sinh(A) Hyperbolic cosecant
date date() Current date string
dos unix_g Execute a UNIX command and return result
eig spec ; bdiag Find eigenvalues and eigenvectors
eval evstr ; execstr Execute a string containing an instruction
fclose mclose Close one or more open files
feof meof Test for end-of-file
ferror mclearerr ; merror Query about errors in file input or output
fft(A[,...]) fft(A,-1[,...]) Discrete Fourier transform
fgetl mgetl Read line(s) from file, discard newline character
fgets fgetstr Read line from file, keep newline character
fliplr(A) A(:,$:-1:1) Flip matrix in left/right direction
flipud(A) A($:-1:1,:) Flip matrix in up/down direction
fopen mopen Open a file or obtain information about open files
frewind(fid) mseek("0",fid) Move the file position indicator to the beginning of an open file
fseek mseek Set file position indicator
ftell mtell Get file position indicator
hankel hank Hankel matrix
ifft(A[,...]) fft(A,1[,...]) Inverse discrete Fourier transform
iscell(A) typeof(A)=="ce" Determine if input is a cell array
ischar(A) type(A)==10 Determine if item is a character array
ishandle(A) type(A)==9 Determines if values are valid graphics object handles
isinteger(A) type(A)==8 Detect whether an array has integer data type
isscalar(A) sum(length(A))==1
isstr(A) type(A)==10 Determine if item is a character array
isstruct(A) typeof(A)=="st" Determine if input is a structure array
isunix getos() "Windows" Determine if Unix version
ispc (getos() == 'Windows') Determine if PC (Windows) version
kron(A,B) A .*. B Kronecker tensor product
lookfor apropos Search for specified keyword in all help entries
lower(str) convstr(str,"l") Convert string to lower case
mod pmodulo Modulus after division
nargin argn(2) Number of function input arguments
nargout argn(1) Number of function output arguments
null kernel Null space of a matrix
num2str string Number to string conversion
ones(size(A)) ones(A) Create an array of all ones
otherwise else Default part of switch/select statement
pause xpause Halt execution temporarily
prod(A,1) prod(A,"r") Product of array elements
rand(A) rand(A[,"uniform"]) Uniformly distributed random numbers and arrays
randn(A) rand(A,"normal") Normally distributed random numbers and arrays
realmax number_properties("huge") Largest positive floating-point number
realmin number_properties("tiny") Smallest positive floating-point number
rem(X,Y) X-fix(X./Y).*Y Remainder after division
reshape matrix Reshape array
strcmp(str1,str2) str1==str2 Compare strings
strfind strindex Find one string within another
strrep strsubst String search and replace
switch select Switch among several cases based on expression
tic tic() Starts a stopwatch timer
toc toc() Read the stopwatch timer
unix unix_g Execute a UNIX command and return result
upper(str) convstr(str,"u") Convert string to upper case
end (index) $ Last index
eps %eps Floating-point relative accuracy
i ; j %i Imaginary unit
pi %pi Ratio of a circle’s circumference to its diameter
Published in: on 14/08/2015 at 21:49  Leave a Comment  

Our homemade 40/80 meter antenna

Recently, we made an antenna for 40 and 80 meter bands in our radio club, UN9GWA.

Here’s the video (in Russian but radio amateurs will understand what is going on):

The antenna is a closed loop of copper wire, a little bit longer than 80 meters.

The exact length was adjusted so that the internal tuner of our Icom IC-7600 was able to tune it for the 40 and 80 meter bands.

The antenna is stretched between our balcony on the ninth floor and the roofs of two neighboring buildings.

Because the wire is thin, the bandwidth is not large, and the antenna is not tuned for the whole 80 m band.

However, its length is adjustable. We soldered several terminals, half a meter apart, so we can change the length which shifts the resonant frequency within the band. This is what we do on the balcony in the video during our antenna party where we also made a fruit salad and cookies with marshmallow and chocolate by melting them under the sun.

Making antennas is fun. Seeing them work is even more fun!

Best wishes from our radio club!

73 de UN9GWA

Published in: on 04/08/2015 at 19:44  Leave a Comment  

Logistic regression in Scilab

Let’s create some random data that are split into two different classes, ‘class 0’ and ‘class 1’.

We will use these data as a training set for logistic regression.

lreg01


b0 = 10;
t = b0 * rand(100,2);
t = [t 0.5+0.5*sign(t(:,2)+t(:,1)-b0)];

b = 1;
flip = find(abs(t(:,2)+t(:,1)-b0)<b);
t(flip,$)=grand(length(t(flip,$)),1,"uin",0,1);

t0 = t(find(t(:,$)==0),:);
t1 = t(find(t(:,$)==1),:);

clf(0);scf(0);
plot(t0(:,1),t0(:,2),'bo')
plot(t1(:,1),t1(:,2),'rx')

The data from different classes overlap slightly. The degree of overlapping is controlled by the parameter b in the code.

We want to build a classification model that estimates the probability that a new, incoming data belong to the class 1.

First, we separate the data into features and results:


x = t(:, 1:$-1); y = t(:, $);

[m, n] = size(x);

Then, we add the intercept column to the feature matrix


// Add intercept term to x
x = [ones(m, 1) x];

The logistic regression hypothesis is defined as:

h(θ, x) = 1 / (1 + exp(−θTx) )

It’s value is the probability that the data with the features x belong to the class 1.

The cost function in logistic regression is

J = [−yT log(h) − (1−y)T log(1−h)]/m

where log is the “element-wise” logarithm, not a matrix logarithm.

If we use the gradient descent algorithm, then the update rule for the θ is

θθαJ = θα xT (hy) / m

The code is as follows


// Initialize fitting parameters
theta = zeros(n + 1, 1);

// Learning rate and number of iterations

a = 0.01;
n_iter = 10000;

for iter = 1:n_iter do
    z = x * theta;
    h = ones(z) ./ (1+exp(-z));
    theta = theta - a * x' *(h-y) / m;
    J(iter) = (-y' * log(h) - (1-y)' * log(1-h))/m;
end

Now, the classification can be visualized:


// Display the result

disp(theta)

u = linspace(min(x(:,2)),max(x(:,2)));

clf(1);scf(1);
plot(t0(:,1),t0(:,2),'bo')
plot(t1(:,1),t1(:,2),'rx')
plot(u,-(theta(1)+theta(2)*u)/theta(3),'-g')

lreg02

Looks good.

The graph of the cost at each iteration is:


// Plot the convergence graph

clf(2);scf(2);
plot(1:n_iter, J');
xtitle('Convergence','Iterations','Cost')

lreg03

Published in: on 04/08/2015 at 18:47  Leave a Comment  
Tags: