I am using LibSVM to carry out some multi-class classifications. I trained the model using the MATLAB interface of LibSVM. I then saved this model in a format that would be recognized in C. I now want to classify using svm_predict in C. I am having trouble being able to reproduce the results that I saw in MATLAB. In fact I get the same class output irrespective of what test vector I feed in (even a vector of zeros) I think the issue is with the way I am loading the test vector x into the svm_node structure. Below is the code snippet. Do let me know if this is correct way or if I am missing something.
struct svm_model *libsvm_model = svm_load_model('mymodel.svm');
struct svm_node x[2001]; // this is for one feature vector of size 2000x1
int index = 1;
int i = 0;
for (i = 0; i < features.size(); i++) {
x[i].index = index;
x[i].value = features.at(i);
index = index + 1;
}
x[i+1].index = -1;
x[i+1].value = '?';
double result = svm_predict(libsvm_model, x);
This seems to be a problem:
x[i+1].index = -1;
x[i+1].value = '?';
libsvm requires svm_node to be an input vector, which should have positive indexes, and double values. You should not "leave" some weird empty dimension.
And by the way, you don't need index variable
for (i = 0; i < features.size(); i++) {
x[i].index = index;
x[i].value = features.at(i);
index = index + 1;
}
is equivalent to
for (i = 0; i < features.size(); i++) {
x[i].index = i + 1;
x[i].value = features.at(i);
}
Related
monthly->maxTemperature = yearData[i].high;
monthly->minTemperature = yearData[i].low;
I just can't seem to understand the logic of what the iterations will look like or how to access the proper elements in the array of data to get the proper data for each month.... without corrupting data. Thanks!
You're on the right track:
void stats(int mth, const struct Data yearData[], int size, struct Monthly* monthStats)
{
// These are used to calc averages
int highSum = 0;
int lowSum = 0;
int days = 0;
// Initialize data
monthly->maxTemperature = INT_MIN;
monthly->minTemperature = INT_MAX;
monthly->totalPrecip = 0;
for (int i = 0; i < size; ++i) {
// Only use data from given month
if (yearData[i].month == mth) {
days += 1;
if (yearData[i].high > monthly->maxTemperature) monthly->maxTemperature = yearData[i].high;
if (yearData[i].low < monthly->minTemperature) monthly->minTemperature = yearData[i].low;
highSum += yearData[i].high;
lowSum + yearData[i].low;
monthly->totalPrecip += yearData[i].precip;
}
}
if (0 != days) {
monthly->avgHigh = highSum / days;
monthly->avgLow = lowSum / days;
}
}
Before working on the assignment it's a good idea to examine the API that you need to implement for clues. First thing to notice is that the reason the struct Monthly is passed to your function by pointer is so that you could set the result into it. This is different from the reason for passing struct Data as a pointer*, which is to pass an array using the only mechanism for passing arrays available in C. const qualifier is a strong indication that you must not be trying to modify anything off of the yearData, only the monthStats.
This tells you what to do with the min, max, average, and total that you are going to find in your function: these need to be assigned to fields of monthStats, like this:
monthStats->maxTemperature = maxTemperature;
monthStats->minTemperature = minTemperature;
...
where maxTemperature, minTemperature, and so on are local variables that you declare before entering the for loop.
As far as the for loop goes, your problem is that you ignore the mth variable completely. You need to use its value to decide if an element of yearData should be considered for your computations or not. The simplest way is to add an if to your for loop:
int maxTemperature = INT_MIN; // you need to include <limits.h>
int minTemperature = INT_MAX; // to get definitions of INT_MIN and INT_MAX
for(int i = 0; i<size; ++i) {
if (yearData[i].month < mth) continue;
if (yearData[i].month > mth) break;
... // Do your computations here
}
* Even though it looks like an array, it is still passed as a pointer
I'm actually trying to figured out how i can iterate through some objects in a for loop with names like "Object1, Object2, Object3..."
Here is a code to exemplify what i'm trying to do:
for(int i = 0; i <= numberOfObjects; i++) {
someVariable = Object1.value/Object2.value/Object3.value;
}
In this case in the part of the code that i have the Object1,Object2 and Object3 i will change for something, example Object[i].
In this case it can't be done with arrays since i don't know how much objects were created.
For the C++ language
Put the objects into a std::vector.
Your loop will look like:
for (unsigned int i = 0; i < object_vector.size(); ++i)
{
result = result / object_vector[i].value;
}
For the C language
object my_objects[] = {object1, object2, /*...*/};
array_size = sizeof(my_objects) / sizeof(object);
for (unsigned int i = 0; i < array_size; ++i)
{
result = result / my_objects[i].value;
}
Use an array:
std::vector<T> objects{Object1, Object2, Object3};
for (auto&& x : objects) {
someVariable = x.value;
}
There's no other way to do it.
I'm trying to make this function create X number of variables using an array. I know that this is technically wrong because I need a constant as my array's value (currently 'x'), but excluding that, what am I missing? Looked at so many code samples and can't figure it out, but I know it's got to be simple...
void variables()
{
int i;
int bars = 10;
int x = 1;
for (i = 1; i <= bars+1; i++)
{
int variables[bars] = { x };
x++;
if (i >= bars+1)
{
break;
}
}
void variables()
{
int bars = 10;
if(bars >= Bars) bars = Bars - 1;
// to be able to set array size based on variable,
// make a dynamically sized array
double highvalues[];
ArrayResize(highvalues, bars);
for (int i = 0 /*Note: Array index is zero-based, 0 is first*/; i <= bars; i++)
{
highvalues[i] = iHigh(NULL, 0, i);
// or
highvalues[i] = High[i];
}
}
It is hard to tell what do you want to achieve.
If you want to fill an array with a value ArrayFill() fill help you.
I want to save a ppm file as a P6. That mean bytes only. Also I want to save it in parallel way, so this is why I am using mmap to map memory.
This is a part of code where I am trying to save everything:
char* map;
//...
int offset = sprintf(map,"P6\n%d %d\n%d\n",x,y,k);
int counter = offset;
for(i = 0; i < x; i++)
{
for(j = 0; j < y; j++)
{
map[counter] = outputRed[i][j];
map[counter+1] = outputGreen[i][j];
map[counter+2] = outputBlue[i][j];
counter++;
}
}
The problem is that I am unable to save those values from Red, green and blue array (all are int) into map values.
Earlier I've simply used fputc with fopen(FILE,"wb") and that was doing everything nice, but now I cannot do it like this.
Can someone tell me how I can do it?
You have to increment the counter by 3 for each pixel:
map[counter] = outputRed[i][j];
map[counter+1] = outputGreen[i][j];
map[counter+2] = outputBlue[i][j];
counter += 3;
You can increment the counter after copying data for each pixel. This may be more easy to understand.
map[counter++] = outputRed[i][j];
map[counter++] = outputGreen[i][j];
map[counter++] = outputBlue[i][j];
Make sure the enough memory is allocated to map.
I'm currently working with the FFTW2 (Fastest Fourier Transform in the West) library, and after writing a successful routine in Fortran, I am moving on to C. However, I'm having some trouble with data assigning when I try to input the data to transform (ie, the values of sin(x)). Currently, my code is:
#include <stdio.h>
#include <fftw.h>
#include <math.h>
#include <complex.h>
void compute_fft(){
int i;
const int n[8];
fftw_complex in[8];
fftwnd_plan p;
in[0]->re = 0;
in[1]->re = (sqrt(2)/2);
in[2]->re = 1;
in[3]->re = (sqrt(2)/2);
in[4]->re = 0;
in[5]->re = -(sqrt(2)/2);
in[6]->re = -1;
in[7]->re = -(sqrt(2)/2);
for(i = 0; i < 8; i++){
(in[i])->im = 0;
}
p = fftwnd_create_plan(8, n, FFTW_FORWARD, FFTW_ESIMATE | FFTW_IN_PLACE);
fftwnd_one(p, &in[0], NULL);
fftwnd_destroy_plan(p);
printf("Sin\n");
for(i = 0; i < 8; i++){
printf("%d\n", n[i]);
}
}
I've been using this http://fftw.org/fftw2_doc/ link for documentation/tutorial purposes, and currently my error is "invalid type argument of a->a (have afftw_complexa)", and the 'a' characters on either side of the -> and fftw_complex have carats above them. I'm using basically the same pattern that worked when I wrote this in Fortran, and I seem to be following the tutorial, it's really just this assignment here that I am messing up the syntax on. For the record, the compiler I am using is nvcc, if it makes a difference (I previously tried gcc). If anyone here has every used FFTW2 in C before, could you help correct my mistake? Thank you!
It might be because of your array "in" which is a array of fftw_complex so instead of using in[0]->re = 0 you should use it as in[0].re = 0. Unless fftw_complex is again typedefined to some array.
fftw_complex in[8];
in[0].re = 0;
in[1].re = (sqrt(2)/2);
in[2].re = 1;
in[3].re = (sqrt(2)/2);
in[4].re = 0;
in[5].re = -(sqrt(2)/2);
in[6].re = -1;
in[7].re = -(sqrt(2)/2);
Since ffwt_complex is an double[2] being the first dimension([0]) for the real data and the second ([1]) for the imaginary data, a safe solution is:
in[0][0] = 0;
in[1][0] = (sqrt(2)/2);
in[2][0] = 1;
in[3][0] = (sqrt(2)/2);
in[4][0] = 0;
in[5][0] = -(sqrt(2)/2);
in[6][0] = -1;
in[7][0] = -(sqrt(2)/2);
for(i = 0; i < 8; i++){
in[i][1] = 0;
}