kronyx 1.1.0

A lightweight deep learning framework built from scratch using NumPy

Kronyx

Built from Scratch • NumPy Powered • Educational Deep Learning Framework

A lightweight deep learning framework that makes neural networks easy to learn, inspect, visualize, and build.

---

Why Kronyx?

Most deep learning frameworks are designed primarily for production.

Kronyx is designed to help you understand deep learning.

Whether you're learning neural networks, teaching machine learning, prototyping new ideas, or building lightweight AI applications, Kronyx provides a familiar Keras-like API with tools that make every layer transparent.

What makes Kronyx different?

• Built entirely from NumPy
• Designed for education first
• 📈 Built-in training visualizations (history.plot())
• Rich model inspection (model.summary())
• Architecture visualization (model.visualize())
• Native .krx model serialization
• ⚡ Lightweight with zero heavyweight ML dependencies
• Familiar Sequential API inspired by Keras
• 📦 Installable directly from PyPI

---

🚀 Installation

pip install kronyx

For development:

git clone https://github.com/BillyMuthiani/Kronyx.git

cd Kronyx

pip install -e ".[dev]"

For plotting support:

pip install kronyx[plotting]

For DataFrame support:

pip install kronyx[dataframe]

---

⚡ Quick Example

import numpy as np
from kronyx import *

X = np.array([
    [0,0],
    [0,1],
    [1,0],
    [1,1]
])

y = np.array([
    [0],
    [1],
    [1],
    [0]
])

model = Sequential()

model.add(Dense(2,16))
model.add(ReLU())

model.add(Dense(16,1))
model.add(Sigmoid())

model.compile(
    loss=BinaryCrossEntropy(),
    optimizer=Adam(),
    metric=Accuracy()
)

history = model.fit(X,y,epochs=1000)

history.plot()

model.summary()

model.visualize()

model.save("xor.krx")

---

Features

Category	Features
Core	Pure NumPy backend, Keras-like Sequential API
Education	Built for learning neural networks, model.summary(), model.visualize()
Visualization	history.plot(), plot_decision_boundary(), plot_confusion_matrix(), plot_dataset(), plot_predictions(), plot_feature_space()
Optimization	SGD, Adam
Losses	BinaryCrossEntropy, CategoricalCrossEntropy, SoftmaxCategoricalCrossEntropy
Metrics	Accuracy, BinaryAccuracy, CategoricalAccuracy, Precision, Recall, F1Score, ConfusionMatrix, TopKAccuracy
Callbacks	EarlyStopping, CSVLogger, ModelCheckpoint, ReduceLROnPlateau
Preprocessing	StandardScaler, MinMaxScaler, RobustScaler, OneHotEncoder
Datasets	xor, spiral, circles, moons, blobs, iris
Data Utilities	train_test_split, BatchLoader, TensorDataset
Serialization	.krx format
Quality	mypy type checking, Ruff linting
Distribution	PyPI

---

📚 Documentation

Guide	Description
Getting Started	Installation and first model
Sequential API	Building models
Layers	Dense, Conv2D, Flatten, Dropout
Callbacks	EarlyStopping, CSVLogger
Serialization	Saving and loading .krx models
Examples	Complete working examples
Roadmap	Future development
Changelog	Release history

🌟 Vision

Kronyx exists to make deep learning understandable.

Instead of treating neural networks as black boxes, Kronyx exposes every layer, every parameter, and every training step through intuitive visualization and inspection tools.

Our mission is to become the best framework for learning how deep learning works under the hood.

---

Made with ❤️ using NumPy.

⭐ If Kronyx helps you learn or build, consider giving the repository a star.