WIP: skeleton for RBF

applications/mixed-precision/dnn/fit_sin.cpp

@@ -82,28 +82,11 @@ void SinFunctionFit() {
 	std::cout << "Result : y = " << a << " + " << b << "x + " << c << "x^2 + " << d << "x^3\n";
 }
 
-// Regression testing guards: typically set by the cmake configuration, but MANUAL_TESTING is an override
-#define MANUAL_TESTING 0
-// REGRESSION_LEVEL_OVERRIDE is set by the cmake file to drive a specific regression intensity
-// It is the responsibility of the regression test to organize the tests in a quartile progression.
-//#undef REGRESSION_LEVEL_OVERRIDE
-#ifndef REGRESSION_LEVEL_OVERRIDE
-#undef REGRESSION_LEVEL_1
-#undef REGRESSION_LEVEL_2
-#undef REGRESSION_LEVEL_3
-#undef REGRESSION_LEVEL_4
-#define REGRESSION_LEVEL_1 1
-#define REGRESSION_LEVEL_2 1
-#define REGRESSION_LEVEL_3 1
-#define REGRESSION_LEVEL_4 1
-#endif
-
 int main()
 try {
 	using namespace sw::universal;
 
 
-#if MANUAL_TESTING
 	using bf16 = cfloat<16, 8, uint16_t, true, true, false>;
 	SinFunctionFit<float>();	// Result : y = -0.2031700 + 0.800356x + -0.0207303x^2 + -0.0852961x^3:  loss = 13.1245
 	SinFunctionFit<fp32>();	    // Result : y = -0.2031700 + 0.800356x + -0.0207303x^2 + -0.0852961x^3:  loss = 13.1245
@@ -129,25 +112,7 @@ try {
 	SinFunctionFit<l16_12>();   // Result : y =  0.1230070 + 0.434995x +  0.5860130x^2 +  0.2949960x^3:  loss = 15.9973
 	SinFunctionFit<l16_14>();   // Result : y =  0         + 0x        +  0.585988x^2  +  0x^3        :  loss = 1.99992
 
-#else
-
-#if REGRESSION_LEVEL_1
-	SinFunctionFit<float>();	// Result : y = -0.2031700 + 0.800356x + -0.0207303x^2 + -0.0852961x^3:  loss = 13.1245
-#endif
-
-#if REGRESSION_LEVEL_2
-
-#endif
-
-#if REGRESSION_LEVEL_3
-
-#endif
-
-#if REGRESSION_LEVEL_4
-
-#endif
 
-#endif
 	return EXIT_SUCCESS;
 }
 catch (char const* msg) {

applications/mixed-precision/dnn/rbf.cpp

@@ -0,0 +1,49 @@
+// Copyright (C) 2017 Stillwater Supercomputing, Inc.
+// SPDX-License-Identifier: MIT
+//
+// This file is part of the universal numbers project, which is released under an MIT Open Source license.
+#include <universal/utility/directives.hpp>
+
+#include <universal/dnn/dnn.hpp>
+
+/*
+Image Preprocessing Precondition:
+ - Load images and convert them to a suitable format (e.g., grayscale or feature vectors).
+ - Normalize pixel values to a common range (e.g., 0-1).
+
+RBF Network Training Steps:
+ - Center Selection: 
+	Choose a subset of training images as centers for the RBF units.
+ - Width Parameter: 
+        Determine the width parameter (sigma) for the RBF functions, controlling their influence.
+ - Weight Training: 
+	Use a supervised learning algorithm (e.g., gradient descent) to adjust the weights connecting 
+	the RBF layer to the output layer.
+
+Image Classification Postcondition:
+  - For a new image, calculate the activation of each RBF unit based on its distance from the centers.
+  - Feed the activations to the output layer and use a decision rule (e.g., maximum activation) to determine the class.
+*/
+
+
+#include <vector>
+#include <cmath>
+
+
+int main() {
+    // Load and preprocess images
+    // ...
+
+
+
+    // Create RBF network
+
+
+    // Train the network
+
+
+    // Test the network
+
+
+    return 0;
+}

include/universal/dnn/dnn.hpp

@@ -3,7 +3,8 @@
 // Super-simple DNN implementation to aid the application,
 // numerical, and reproducibility examples.
 //
-// Copyright (C) 2021 Stillwater Supercomputing, Inc.
+// Copyright (C) 2017 Stillwater Supercomputing, Inc.
+// SPDX-License-Identifier: MIT
 //
 // This file is part of the universal numbers project, which is released under an MIT Open Source license.
 #ifndef _UNIVERSAL_DNN_LIBRARY