[feat][nmsis] add nmsis component and nn,dsp demo

examples/dsp/SVMFunctions_polynomialSVM/main.c

@@ -0,0 +1,118 @@
+#include "riscv_math.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include "../common.h"
+
+#include "../HelperFunctions/math_helper.c"
+#include "../HelperFunctions/ref_helper.c"
+
+#include <stdio.h>
+#define DELTAF32 (0.05f)
+#define DELTAQ31 (63)
+#define DELTAQ15 (1)
+#define DELTAQ7 (1)
+
+int test_flag_error = 0;
+
+/* 
+  The polynomial SVM instance containing all parameters.
+  Those parameters can be generated with the python library scikit-learn.
+ */
+riscv_svm_polynomial_instance_f32 params;
+
+/*
+  Parameters generated by a training of the SVM classifier
+  using scikit-learn and some random input data.
+ */
+#define NB_SUPPORT_VECTORS 9
+
+/*
+  Dimension of the vector space. A vector is your feature.
+  It could, for instance, be the pixels of a picture or the FFT of a signal.
+ */
+#define VECTOR_DIMENSION 2
+
+const float32_t dualCoefficients[NB_SUPPORT_VECTORS]={-0.05355985f,-0.08799571f,0.03355538f,0.03568506f,0.02952993f,0.00270412f,0.01743034f,0.01808445f,0.00456627f}; /* Dual coefficients */
+
+const float32_t supportVectors[NB_SUPPORT_VECTORS*VECTOR_DIMENSION]={1.35662086f,0.39208881f,-1.01276844f,-1.13381492f,2.18118916f,1.76274021f,-2.62629126f,-0.81061583f,-0.86876703f,-2.74531146f,-0.82722208f,2.77116307f,-1.51435674f,2.40297933f,2.61838172f,-1.22023814f,2.0558457f,-2.00230947f}; /* Support vectors */
+
+/*
+  Class A is identified with value 0.
+  Class B is identified with value 1.
+  
+  This array is used by the SVM functions to do a conversion and ease the comparison
+  with the Python code where different values could be used.
+ */
+const int32_t   classes[2]={0,1};
+
+int main()
+{
+    int i;
+    BENCH_INIT;
+    /* Array of input data */
+    float32_t in[VECTOR_DIMENSION];
+
+    /* Result of the classifier */
+    int32_t result;
+    
+
+    /*
+        Initialization of the SVM instance parameters.
+        Additional parameters (intercept, degree, coef0 and gamma) are also coming from Python.
+    */
+    riscv_svm_polynomial_init_f32(&params,
+        NB_SUPPORT_VECTORS,
+        VECTOR_DIMENSION,
+        -1.704476f,        /* Intercept */
+        dualCoefficients,
+        supportVectors,
+        classes,
+        3,                 /* degree */
+        1.100000f,         /* Coef0 */
+        0.500000f          /* Gamma */
+    );
+
+
+    /*
+        Input data.
+        It is corresponding to a point inside the first class.
+    */
+    in[0] = 0.4f;
+    in[1] = 0.1f;
+    BENCH_START(riscv_svm_polynomial_predict_f32);
+    riscv_svm_polynomial_predict_f32(&params, in, &result);
+    BENCH_END(riscv_svm_polynomial_predict_f32);
+    /* Result should be 0 : First class */
+    if (result != 0) {
+        BENCH_ERROR(riscv_svm_polynomial_predict_f32);
+        printf("expect: %d, actual: %d\n", 0, result);
+        test_flag_error = 1;
+    }
+    BENCH_STATUS(riscv_svm_polynomial_predict_f32);
+
+    /*
+        This input vector is corresponding to a point inside the second class.
+    */
+    in[0] = 3.0f;
+    in[1] = 0.0f;
+
+    riscv_svm_polynomial_predict_f32(&params, in, &result);
+
+    /* Result should be 0 : First class */
+    if (result != 1) {
+        BENCH_ERROR(riscv_svm_polynomial_predict_f32);
+        printf("expect: %d, actual: %d\n", 1, result);
+        test_flag_error = 1;
+    }
+    BENCH_STATUS(riscv_svm_polynomial_predict_f32);
+
+
+    BENCH_FINISH;
+    if (test_flag_error) {
+        printf("test error apprears, please recheck.\n");
+        return 1;
+    } else {
+        printf("all test are passed. Well done!\n");
+    }
+    return 0;
+};