maten 0.1.0

Pure Python scientific computing — replaces NumPy, Pandas, Matplotlib & TensorFlow

Maten

A lightweight, dependency-free Python toolkit that provides NumPy-, Pandas-, Matplotlib-, and TensorFlow-like APIs for learning and prototyping.

Features

• maten.core — N-dimensional tensors, element-wise math, and linear algebra
• maten.frame — Series, DataFrame, groupby, and CSV/JSON I/O
• maten.plot — SVG charts (line, bar, scatter, hist, pie, heatmap)
• maten.neural — Scalar autograd, layers, optimizers, and a simple trainer

Installation

From the project root:

pip install .

Editable install for development:

pip install -e ".[dev]"

Quick start

import maten
from maten.core import Tensor, zeros
from maten.frame import DataFrame
from maten.plot import subplots, line
from maten.neural import Value

print(maten.__version__)
print(zeros(3))
print(DataFrame({"x": [1, 2, 3]}))

_, ax = subplots()
line(ax, [0, 1, 2], [0, 1, 4])
ax.save("demo.svg")

x = Value(2.0)
y = x * x
y.backward()
print(x.grad)  # 4.0

Examples

python examples/core_demo.py
python examples/frame_demo.py
python examples/plot_demo.py
python examples/neural_demo.py

Tests

pytest

Documentation

• Core
• Frame
• Plot
• Neural

License

MIT — see LICENSE.