The zeta-learn from dee-z

zeta-learn

zeta-learn is a minimalistic python machine learning library designed to deliver fast and easy model prototyping.

zeta-learn aims to provide an extensive understanding of machine learning through the use of straightforward algorithms and readily implemented examples making it a useful resource for researchers and students.

Documentation: https://zeta-learn.com

Python versions: 3.5 and above

Free software: MIT license

Dependencies

numpy >= 1.15.0

matplotlib >= 2.0.0

Features

Keras like Sequential API for building models.

Built on Numpy and Matplotlib.

Examples folder with readily implemented machine learning models.

Install

pip install ztlearn

Examples

Principal Component Analysis (PCA)

DIGITS Dataset - PCA

MNIST Dataset - PCA

KMEANS

K-Means Clustering (4 Clusters)

Convolutional Neural Network (CNN)

DIGITS Dataset Model Summary

DIGITS CNN

Input Shape: (1, 8, 8)
+---------------------+---------+--------------+
¦ LAYER TYPE          ¦  PARAMS ¦ OUTPUT SHAPE ¦
+---------------------+---------+--------------+
¦ Conv2D              ¦     320 ¦   (32, 8, 8) ¦
¦ Activation: RELU    ¦       0 ¦   (32, 8, 8) ¦
¦ Dropout             ¦       0 ¦   (32, 8, 8) ¦
¦ BatchNormalization  ¦   4,096 ¦   (32, 8, 8) ¦
¦ Conv2D              ¦  18,496 ¦   (64, 8, 8) ¦
¦ Activation: RELU    ¦       0 ¦   (64, 8, 8) ¦
¦ MaxPooling2D        ¦       0 ¦   (64, 7, 7) ¦
¦ Dropout             ¦       0 ¦   (64, 7, 7) ¦
¦ BatchNormalization  ¦   6,272 ¦   (64, 7, 7) ¦
¦ Flatten             ¦       0 ¦     (3,136,) ¦
¦ Dense               ¦ 803,072 ¦       (256,) ¦
¦ Activation: RELU    ¦       0 ¦       (256,) ¦
¦ Dropout             ¦       0 ¦       (256,) ¦
¦ BatchNormalization  ¦     512 ¦       (256,) ¦
¦ Dense               ¦   2,570 ¦        (10,) ¦
+---------------------+---------+--------------+

TOTAL PARAMETERS: 835,338

DIGITS Dataset Model Results

DIGITS Dataset Model Loss

DIGITS Dataset Model Accuracy

MNIST Dataset Model Summary

MNIST CNN

Input Shape: (1, 28, 28)
+---------------------+------------+--------------+
¦ LAYER TYPE          ¦     PARAMS ¦ OUTPUT SHAPE ¦
+---------------------+------------+--------------+
¦ Conv2D              ¦        320 ¦ (32, 28, 28) ¦
¦ Activation: RELU    ¦          0 ¦ (32, 28, 28) ¦
¦ Dropout             ¦          0 ¦ (32, 28, 28) ¦
¦ BatchNormalization  ¦     50,176 ¦ (32, 28, 28) ¦
¦ Conv2D              ¦     18,496 ¦ (64, 28, 28) ¦
¦ Activation: RELU    ¦          0 ¦ (64, 28, 28) ¦
¦ MaxPooling2D        ¦          0 ¦ (64, 27, 27) ¦
¦ Dropout             ¦          0 ¦ (64, 27, 27) ¦
¦ BatchNormalization  ¦     93,312 ¦ (64, 27, 27) ¦
¦ Flatten             ¦          0 ¦    (46,656,) ¦
¦ Dense               ¦ 11,944,192 ¦       (256,) ¦
¦ Activation: RELU    ¦          0 ¦       (256,) ¦
¦ Dropout             ¦          0 ¦       (256,) ¦
¦ BatchNormalization  ¦        512 ¦       (256,) ¦
¦ Dense               ¦      2,570 ¦        (10,) ¦
+---------------------+------------+--------------+

TOTAL PARAMETERS: 12,109,578