In K-fold cross-validation, the aim is to generate K training/validation set pair, where training and validation sets on fold i do no overlap. First, we divide the dataset X into K parts as X1; X2; ... ; XK. Then for each fold i, we use Xi as the validation set and the remaining as the training set.
Possible values of K are 10 or 30. One extreme case of K-fold cross-validation is leave-one-out, where K = N and each validation set has only one instance. If we have more computation power, we can have multiple runs of K-fold cross-validation, such as 10 x 10 cross-validation or 5 x 2 cross-validation.
If we have very small datasets, we do not insist on the non-overlap of training and validation sets. In bootstrapping, we generate K multiple training sets, where each training set contains N examples (like the original dataset). To get N examples, we draw examples with replacement. For the validation set, we use the original dataset. The drawback of bootstrapping is that the bootstrap samples overlap more than the cross-validation sample, hence they are more dependent.
You can also see Java, Python, Cython, Swift, Js, Php, C#, or C++ repository.
- C Editor
- Git
Install the latest version of Git.
In order to work on code, create a fork from GitHub page. Use Git for cloning the code to your local or below line for Ubuntu:
git clone <your-fork-git-link>
A directory called Sampling-C will be created. Or you can use below link for exploring the code:
git clone https://github.com/starlangsoftware/Sampling-C.git
To import projects from Git with version control:
-
Open Rider IDE, select Get From Version Control.
-
In the Import window, click URL tab and paste github URL.
-
Click open as Project.
Result: The imported project is listed in the Project Explorer view and files are loaded.
From IDE
After being done with the downloading and opening project, select Build Solution option from Build menu. After compilation process, user can run Sampling-CS.
k. eğitim kümesini elde etmek için
ArrayList<T> GetTrainFold(int k)
k. test kümesini elde etmek için
ArrayList<T> GetTestFold(int k)
Bootstrap için BootStrap sınıfı
Bootstrap(List<T> instanceList, int seed)
Örneğin elimizdeki veriler a adlı ArrayList'te olsun. Bu veriler üstünden bir bootstrap örneklemi tanımlamak için (5 burada rasgelelik getiren seed'i göstermektedir. 5 değiştirilerek farklı samplelar elde edilebilir)
bootstrap = Bootstrap(a, 5);
ardından üretilen sample'ı çekmek için ise
sample = bootstrap.getSample();
yazılır.
K kat çapraz geçerleme için KFoldCrossValidation sınıfı
KFoldCrossValidation(List<T> instanceList, int K, int seed)
Örneğin elimizdeki veriler a adlı ArrayList'te olsun. Bu veriler üstünden 10 kat çapraz geçerleme yapmak için (2 burada rasgelelik getiren seed'i göstermektedir. 2 değiştirilerek farklı samplelar elde edilebilir)
kfold = KFoldCrossValidation(a, 10, 2);
ardından yukarıda belirtilen getTrainFold ve getTestFold metodları ile sırasıyla i. eğitim ve test kümeleri elde edilebilir.
Stratified K kat çapraz geçerleme için StratifiedKFoldCrossValidation sınıfı
StratifiedKFoldCrossValidation(List<T>[] instanceLists, int K, int seed)
Örneğin elimizdeki veriler a adlı ArrayList of listte olsun. Stratified bir çapraz geçerlemede sınıflara ait veriler o sınıfın oranında temsil edildikleri için her bir sınıfa ait verilerin ayrı ayrı ArrayList'te olmaları gerekmektedir. Bu veriler üstünden 30 kat çapraz geçerleme yapmak için (4 burada rasgelelik getiren seed'i göstermektedir. 4 değiştirilerek farklı samplelar elde edilebilir)
stratified = StratifiedKFoldCrossValidation(a, 30, 4);
ardından yukarıda belirtilen getTrainFold ve getTestFold metodları ile sırasıyla i. eğitim ve test kümeleri elde edilebilir.
- First install conan.
pip install conan
Instructions are given in the following page:
https://docs.conan.io/2/installation.html
- Add conan remote 'ozyegin' with IP: 104.247.163.162 with the following command:
conan remote add ozyegin http://104.247.163.162:8081/artifactory/api/conan/conan-local --insert
- Use the comman conan list to check for installed packages. Probably there are no installed packages.
conan list
- Put the correct dependencies in the requires part
requires = ["math_c/1.0.0", "data_structure_c/1.0.0", "classification_c/1.0.0"]
- Default settings are:
settings = "os", "compiler", "build_type", "arch"
options = {"shared": [True, False], "fPIC": [True, False]}
default_options = {"shared": True, "fPIC": True}
exports_sources = "src/*", "Test/*"
def layout(self):
cmake_layout(self, src_folder="src")
def generate(self):
tc = CMakeToolchain(self)
tc.generate()
deps = CMakeDeps(self)
deps.generate()
def build(self):
cmake = CMake(self)
cmake.configure()
cmake.build()
def package(self):
copy(conanfile=self, keep_path=False, src=join(self.source_folder), dst=join(self.package_folder, "include"), pattern="*.h")
copy(conanfile=self, keep_path=False, src=self.build_folder, dst=join(self.package_folder, "lib"), pattern="*.a")
copy(conanfile=self, keep_path=False, src=self.build_folder, dst=join(self.package_folder, "lib"), pattern="*.so")
copy(conanfile=self, keep_path=False, src=self.build_folder, dst=join(self.package_folder, "lib"), pattern="*.dylib")
copy(conanfile=self, keep_path=False, src=self.build_folder, dst=join(self.package_folder, "bin"), pattern="*.dll")
def package_info(self):
self.cpp_info.libs = ["ComputationalGraph"]
- Set the C standard with compiler flags.
set(CMAKE_C_STANDARD 17)
set(CMAKE_C_FLAGS "-O3")
- Dependent packages should be given with find_package.
find_package(util_c REQUIRED)
find_package(data_structure_c REQUIRED)
- For library part, use add_library and target_link_libraries commands. Use m library for math linker in Linux.
add_library(Math src/Distribution.c src/Distribution.h src/DiscreteDistribution.c src/DiscreteDistribution.h src/Vector.c src/Vector.h src/Eigenvector.c src/Eigenvector.h src/Matrix.c src/Matrix.h src/Tensor.c src/Tensor.h)
target_link_libraries(Math util_c::util_c data_structure_c::data_structure_c m)
- For executable tests, use add_executable and target_link_libraries commands. Use m library for math linker in Linux.
add_executable(DiscreteDistributionTest src/Distribution.c src/Distribution.h src/DiscreteDistribution.c src/DiscreteDistribution.h src/Vector.c src/Vector.h src/Eigenvector.c src/Eigenvector.h src/Matrix.c src/Matrix.h src/Tensor.c src/Tensor.h Test/DiscreteDistributionTest.c)
target_link_libraries(DiscreteDistributionTest util_c::util_c data_structure_c::data_structure_c m)
- Add data files to the cmake-build-debug folder.
- Define structs as class counterparts.
typedef struct {
double *data; // Flattened data array
int *shape; // Shape of the tensor
int *strides; // Strides for each dimension
int dimensions; // Number of dimensions
int total_elements;
} Tensor;
- Define also typedefs for pointer to those structs.
typedef Tensor *Tensor_ptr;
- Put all those definitions to the header files with prototypes of functions.
- Constructor method(s) should allocate a pointer of struct and fill the details and return the pointer.
Vector_ptr create_vector(Array_list_ptr values) {
Vector_ptr result = malloc_(sizeof(Vector));
result->values = values;
result->size = values->size;
return result;
}
- Destructor method should be defined for each struct and deallocate the object and its contents if allocated.
void free_vector(Vector_ptr vector) {
free_array_list(vector->values, free_);
free_(vector);
}
- If there is multiple inheritance, then define the super class inside the subclass, which also should be defined inside its own subclass etc.
struct sgd_momentum {
Optimizer optimizer;
Hash_map_ptr velocity_map;
double momentum;
};
struct Adam {
Sgd_momentum sgd;
Hash_map_ptr momentum_map;
double beta2;
double epsilon;
double current_beta1;
double current_beta2;
};
struct AdamW {
Adam adam;
double weight_decay;
};
- If there is overloading, and subclasses should override the method, you can add pointers to the functions and set them in the constructor.
struct function {
Function_type function_type;
Tensor_ptr (*calculate)(const void*, const Tensor*);
Tensor_ptr (*derivative)(const void*, const Tensor*, const Tensor*);
};
Elu_ptr create_elu(const double a) {
Elu_ptr delu = malloc_(sizeof(Elu));
delu->a = a;
delu->function.function_type = ELU;
delu->function.calculate = calculate_elu;
delu->function.derivative = derivative_elu;
return delu;
}
- Instanceof could be done using an enumerated type defined in the parent class.
enum optimizer_type {
OPTIMIZER, ADAM, ADAM_W, SGD_MOMENTUM, SGD,
};
typedef enum optimizer_type Optimizer_type;
struct optimizer {
Optimizer_type type;
- If needed, hash function should be declared for the struct.
unsigned int hash_function_computational_node(const Computational_node *node, int N) {
return ((uintptr_t)node) % N;
}
- If needed, comparator function should be declared for the struct.
int compare_annotated_word(const Annotated_word *word1, const Annotated_word *word2) {
return compare_string(word1->name, word2->name);
}
- Data Structures: For hashmap, linked hash map and hashset, use create_hash_map, create_linked_hash_map, and create_hash_set in data structure library. There are default string and integer versions of hash_map.
Hash_map_ptr create_hash_map(unsigned int (*hash_function)(const void *, int), int (*key_compare)(const void *, const void *));
Linked_hash_map_ptr create_linked_hash_map(unsigned int (*hash_function)(const void *, int), int (*compare)(const void *, const void *));
Hash_map_ptr create_string_hash_map();
Hash_map_ptr create_integer_hash_map();
Hash_set_ptr create_hash_set(unsigned int (*hash_function)(const void *, int), int (*compare)(const void *, const void *));
- Data Structures: For array list, use create_array_list in data structure library..
Array_list_ptr create_array_list();
- Data Structures: For linked list, use linked_list in data structure library..
Linked_list_ptr create_linked_list(int (*compare)(const void *, const void *));
- Data Structures: For queue and stack, use queue and stack in data structure library.
Queue_ptr create_queue();
Stack_ptr create_stack();
- Data Structures: For hash map that is counting items, use counter hash map in data structure library.
Counter_hash_map_ptr
create_counter_hash_map(unsigned int (*hash_function)(const void *, int), int (*key_compare)(const void *, const void *));
- Do not forget to comment each function.
/**
* Initializes a tensor with given data and shape. Data is cloned in the tensor
*
* @param data Flattened array representing the tensor data.
* @param shape Array representing the shape of the tensor.
* @param dimensions Size of the shape array.
* @return Pointer to the created tensor. Returns NULL on failure.
*/
Tensor_ptr create_tensor(const double *data, const int *shape, int dimensions) {
- Function names should follow snake case case.
Tensor_ptr tensor_get(const Tensor *tensor, const int *dimensions, int size) {
- All heap allocations should be done with malloc_, calloc_, realloc_ taken from the memory management library.
int *new_shape = malloc_((tensor->dimensions - size) * sizeof(int));
- All heap deallocations should be done with free_ taken from the memory management library.
free_(new_shape);
- DO NOT write getter and setter methods.
- If there are multiple constructors for a class, define them as constructor1, constructor2, ....
Tensor_ptr create_tensor(const double *data, const int *shape, int dimensions);
Tensor_ptr create_tensor2(const int *shape, int dimensions);
Tensor_ptr create_tensor3(double *data, const int *shape, int dimensions);
Tensor_ptr create_tensor4(const int *shape, int dimensions);
- Use separate main method for testing purposes.
int main() {
start_medium_memory_check();
testColumnWiseNormalize();
testMultiplyWithConstant();
testDivideByConstant();
end_memory_check();
}
- Testing main method should start with memory check and end with memory check.
int main() {
start_medium_memory_check();
testColumnWiseNormalize();
testMultiplyWithConstant();
testDivideByConstant();
end_memory_check();
}
- Enumerated types should be declared with enum.
enum category_type{
MATHEMATICS, SPORT, MUSIC, SLANG, BOTANIC,
PLURAL, MARINE, HISTORY, THEOLOGY, ZOOLOGY,
METAPHOR, PSYCHOLOGY, ASTRONOMY, GEOGRAPHY, GRAMMAR,
MILITARY, PHYSICS, PHILOSOPHY, MEDICAL, THEATER,
ECONOMY, LAW, ANATOMY, GEOMETRY, BUSINESS,
PEDAGOGY, TECHNOLOGY, LOGIC, LITERATURE, CINEMA,
TELEVISION, ARCHITECTURE, TECHNICAL, SOCIOLOGY, BIOLOGY,
CHEMISTRY, GEOLOGY, INFORMATICS, PHYSIOLOGY, METEOROLOGY,
MINERALOGY
};
- Every header file should start with
#ifndef MATH_DISTRIBUTION_H
#define MATH_DISTRIBUTION_H
and end with
#endif //MATH_DISTRIBUTION_H
- Use regular package for regular expression handling.
void check_match(char* expression, char* word){
Regular_expression_ptr re = create_regular_expression(expression);
if (!full_matches(re, word)){
printf("Mismatch %s %s\n", re->expression, word);
}
free_regular_expression(re);
}
- Do not forget to use const expression for parameters, if they will not be changed in the function.
bool frame_lexical_unit_exists(const Frame* frame, const char *synSetId);
- For dynamic string handling, use util package with stringutils.
struct string {
char *s;
int max_size;
};
typedef struct string String;
typedef String *String_ptr;
String_ptr create_string();
String_ptr create_string2(const char *s);
String_ptr create_string3(const char* s1, const char* s2);
String_ptr create_string4(const char* s1, const char* s2, const char* s3);
- Foor boolean type, use bool.
bool check_definition(const Data_set* data_set, const Instance* instance);
- Use xmlparser package for parsing xml files.
Xml_document_ptr doc = create_xml_document("test.xml");