Link™

A JupyterLab extension for creating a pipeline with notebook cells

Usage Environment

Supported Platform	Linux (AMD64, ARM64) / macOS (Intel, Apple Silicon) / Windows
Python Version	3.8, 3.9, 3.10
Requirements	Git

Install

We recommend installing our program in a new virtual environment.

Step 1. Install Link with the following command.

python3 -m pip install --upgrade mrx-link

Step 2. After installation, run JupyterLab.

python3 -m jupyterlab

Features

Pipeline Creation

Create a pipeline with notebook cells to improve code readability and reproducibility.

Caching Management

Store execution results for respective cells to minimize redundant cell executions.

Remote Resources

Optimize the use of server resources by using separate resources for different pipeline components.

Hyper-Parameter Optimizer

Link provides a hyper-parameter feature, which works to find optimal hyper-parameters in an automated manner.

Version Control

Link provides a version control feature, where you can easily view changes to your pipeline and merge conflicts.

Link SDK

Link provides an SDK, where you can easily run the Link notebook file with the command line interface, and create, run, export pipelines with Python code.

Examples with Link Pipelines

• Deep Q-Network Reinforcement Learning for the CartPole Environment

• Denoising Autoencoder with PyTorch

• Iris Classification

• MNIST Image Classification with PyTorch

• MNIST Image Generation with VAE

• Spiral-Distributed Data Classification

• Sequential Data Classification

• Titanic Data EDA and Modeling

• XGBoost Example