bareml 0.0.9

bareml is a set of "bare" implementations of machine learning / deep learning algorithms from scratch (only depending on numpy) in Python. "bare" means to aim at:

1. Code as a direct translation of the algorithm / formula
2. With minimum error handling and efficiency gain tricks

To maximise understandability of the code, interface of modules in bareml/machinelearning/ is aligned to Scikit-learn, and interface of modules in bareml/deeplearning/ is aligned to PyTorch, as seen in below 2 examples.

Example1:

from bareml.machinelearning.utils.model_selection import train_test_split
from bareml.machinelearning.supervised import KernelRidge

# assume the data X, y are defined
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

reg = KernelRidge(alpha=1, kernel='rbf')
reg.fit(X_train, y_train)
y_pred = reg.predict(X_test)
print(reg.score(X_test, y_test))

Example2:

from bareml.deeplearning import layers as nn
from bareml.deeplearning import functions as F

class Net(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=32, kernel_size=3, stride=1)
        self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, stride=1)
        self.dropout1 = nn.Dropout(p=0.25)
        self.dropout2 = nn.Dropout(p=0.5)
        self.fc1 = nn.Linear(in_features=33856, out_features=128)
        self.fc2 = nn.Linear(in_features=128, out_features=10)

    def forward(self, x):
        x = self.conv1(x)
        x = F.relu(x)
        x = self.conv2(x)
        x = F.relu(x)
        x = F.max_pool2d(x, 2)
        x = self.dropout1(x)
        x = x.flatten()
        x = self.fc1(x)
        x = F.relu(x)
        x = self.dropout2(x)
        x = self.fc2(x)
        return x

Installation

$ pip install bareml

or

$ git clone https://github.com/shotahorii/bareml.git
$ cd bareml
$ python setup.py install

Dependencies

Mandatory

• numpy

Optional

• cupy
• PIL
• matplotlib

Examples

Generating handwriting digits by GAN

[Google Colab]

Implementations

Deep Learning

• Pytorch-like Deep Learning Framework

Supervised Learning

• Bernoulli Naive Bayes
• Decision Trees
• Elastic Net
• Gaussian Naive Bayes
• Generalised Linear Model
• K Nearest Neighbors
• Kernel Ridge Regression
• Lasso Regression
• Linear Regression
• Logistic Regression
• Perceptron
• Poisson Regression
• Ridge Regression

Unsupervised Learning

• KMeans (KMeans++)
• DBSCAN

Ensemble Learning

• AdaBoost
• AdaBoost M1
• AdaBoost Samme
• AdaBoost RT
• AdaBoost R2
• Bagging
• Gradient Boosting
• Random Forest
• Stacking
• Voting
• XGBoost

Utilities

• Preprocessing (Scaler, Encoder etc)
• Metrics
• Kernel functions
• Probability Distributions
• Model Selection (KFold etc)

References

• Deep learning programs are based on O'Reilly Japan's book "Deep learning from scratch 3" (Koki Saitoh) and its implementation Dezero.
• References of machine learning programs are documented in each source file, but mostly based on original papers, "Pattern Recognition and Machine Learning" (Christopher M. Bishop) and/or "Machine Learning: A Probabilistic Perspective" (Kevin P. Murphy).