Skip to main content

No project description provided

Project description


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
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
sklearn.preprocessing.StandardScaler clara.transpiler.preprocessing.StandardScalerTranspiler


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()


# 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((,
scale = StandardScaler()

pca = PCA(n_components=0.8)

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

from clara.transpiler.pca import PCATranspiler

transpiler = PCATranspiler(pca)

code = transpiler.generate_code()

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

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);


    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]);


    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*


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((,

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

transpiler = MLPCTranspiler(mlp)

code = transpiler.generate_code()

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


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((,

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

transpiler = MLPRTranspiler(mlp)

code = transpiler.generate_code()

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

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.

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

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for clara-transpiler, version 0.17.5
Filename, size File type Python version Upload date Hashes
Filename, size clara_transpiler-0.17.5-py3-none-any.whl (26.4 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size clara_transpiler-0.17.5.tar.gz (13.7 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page