Skip to content

asergiobranco/clara

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

64 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Clara

CLARA is a tool designed to help Machine Learning developers to build their models using High-Level languages (Python), but easily implement them in C. The goal is not to convert the code, but to convert the trained model itself (the object). Therefore, this is not a code converter, but a code transpiler.

The following algorithms are available

  • Classification
    • MLP
    • Decision Tree
    • Support-Vector Machines (SVC & Nu)
    • LinearSVM
    • Gaussian Naive Bayes
    • Complement Naive Bayes
    • Multinomial Naive Bayes
    • Categorical Naive Bayes
    • Bernoulli Naive Bayes
  • Regression
    • MLP
    • Support-Vector Machines
  • Decomposition
    • PCA
  • Preprocessing
    • StandardScaler
    • KernelCenterer
    • MaxAbsScaler
    • MinMaxScaler
    • RobustScaler

Transpiling Tools

Python Class Clara Class
Decomposition
sklearn.decomposition.PCA clara.transpiler.pca.PCATranspiler
Neural Networks
sklearn.neural_network.MLPClassifier clara.transpiler.mlp.MLPCTranspiler
sklearn.neural_network.MLPRegressor clara.transpiler.mlp.MLPRTranspiler
Decision Tree
sklearn.tree.DecisionTreeClassifier clara.transpiler.tree.DecisionTreeClassifierTranspiler
Support-Vector Machines
sklearn.svm.SVC clara.transpiler.svm.SVCTranspiler
sklearn.svm.NuSVC clara.transpiler.svm.SVCTranspiler
sklearn.svm.LinearSVM clara.transpiler.svm.LinearSVMTranspiler
sklearn.svm.SVR clara.transpiler.svm.SVRTranspiler
Naive Bayes
sklearn.naive_bayes.GaussianNB clara.transpiler.naive_bayes.GaussianNBTranspiler
sklearn.naive_bayes.ComplementNB clara.transpiler.naive_bayes.ComplementNBTranspiler
sklearn.naive_bayes.MultinomialNB clara.transpiler.naive_bayes.MultinomialNBTranspiler
sklearn.naive_bayes.CategoricalNB clara.transpiler.naive_bayes.CategoricalNBTranspiler
sklearn.naive_bayes.BernoulliNB clara.transpiler.naive_bayes.BernoulliNBTranspiler
Preprocessing
sklearn.preprocessing.StandardScaler clara.transpiler.preprocessing.StandardScalerTranspiler

Syntax

Besides the multiple available algorithms, the syntax to use in any of them is the same and shown in the snippet bellow.

#The ML Algorithm you want to use
model = ScikitLearnClass()

# TRAIN YOUR MODEL FIRST!!!
model.fit()

# The correspondent CLARA TRanspiler Class
transpiler = ClaraClassTranspiler(model) #The correspondent Clara Class


# The C code to be exported to a .c file and compiled
# The code is of the model trained, therefore no retraining is needed.
c_code = transpiler.generate_code()

PCA Transpiler

Python Exporting

from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
from sklearn.datasets import load_wine

data = load_wine()
dataset = np.column_stack((data.data, data.target))
scale = StandardScaler()

pca = PCA(n_components=0.8)

X = scale.fit_transform(dataset[::,:-1])
pca.fit(X)

from clara.transpiler.pca import PCATranspiler

transpiler = PCATranspiler(pca)

code = transpiler.generate_code()

with open("pca.c", "w+") as fp:
  fp.write(code)

Test code in C

The results may vary, but if they should be the same!!

int main(int argc, const char * argv[]) {
    // insert code here...
    double sample[N_FEATURES] = { 1.51861254, -0.5622498 ,  0.23205254, -1.16959318,  1.91390522,
        0.80899739,  1.03481896, -0.65956311,  1.22488398,  0.25171685,
                                  0.36217728,  1.84791957,  1.01300893};
    double scores[N_COMPONENTS] = {0};

    double inverse_sample[N_FEATURES] = {0};

    calculate_scores(sample, scores);

    printf("\nScores\n");


    for(int i = 0; i < N_COMPONENTS; i++){
        printf("%f\t", scores[i]);
    }

    printf("\n\nInverse Transform\n");

    inverse(scores, inverse_sample);

    for(int i = 0; i < N_FEATURES; i++){
        printf("%f\t", inverse_sample[i]);
    }

    printf("\n\n");

    pca_dimensions_t val;

    val = calculate_dimensions(sample);

    printf("T2 = %f, Q-Residuals: %f\n\n", val.hoteling2, val.q_residuals);



}

MLP Transpiler

Multi-Layer Perceptron are the basis of Neural Networks and Deep Learning. Our tools provides a way to transpile MLPs for regression and classification problems.

  • Note: At the current time, binary classifications are not working... Sorry*

MLPClassifier

from sklearn.neural_network import MLPClassifier
from sklearn.datasets import load_wine
import numpy as np

from clara.transpiler.mlp import MLPCTranspiler

data = load_wine()
dataset = np.column_stack((data.data, data.target))

mlp = MLPClassifier(hidden_layer_sizes=(30, 10))

mlp.fit(ddataset.data, dataset.target)


transpiler = MLPCTranspiler(mlp)

code = transpiler.generate_code()

with open("mlp.c", "w+") as fp:
  fp.write(code)

MLPRegressor

from sklearn.neural_network import MLPRegressor
from sklearn.datasets import load_boston
import numpy as np

from clara.transpiler.mlp import MLPRTranspiler

data = load_boston()
dataset = np.column_stack((data.data, data.target))

mlp = MLPClassifier(hidden_layer_sizes=(30, 10))

mlp.fit(dataset.data, dataset.target)



transpiler = MLPRTranspiler(mlp)

code = transpiler.generate_code()

with open("mlp.c", "w+") as fp:
  fp.write(code)

Test code in C

int main(){
    double s[N_FEATURES] = {14.23, 1.71, 2.43, 15.6, 127.0, 2.8, 3.06, 0.28, 2.29, 5.64, 1.04, 3.92, 1065.0};
    int class;
    for(int i = 0; i<N_FEATURES; i++){
      sample[i] = s[i];
    }
    class = predict(sample);
    return 0;
}

Cite Us

Please, if you use our tool in any of your projects, cite us. This will help us improve and look at what people may need! Thanks!

DOI: 10.5281/zenodo.3930335

Sérgio Branco. (2020, July 4). CLARA - Embedded ML Tools (Version v0.0.1). Zenodo. http://doi.org/10.5281/zenodo.3930336

@software{sergio_branco_2020_3930336,
  author       = {Sérgio Branco},
  title        = {CLARA - Embedded ML Tools},
  month        = jul,
  year         = 2020,
  publisher    = {Zenodo},
  version      = {v0.0.1},
  doi          = {10.5281/zenodo.3930336},
  url          = {https://doi.org/10.5281/zenodo.3930336}
}