Libra AI
Create a Neural Network in One Line of Code
Published by Sid Chadha and Palash Shah (Creator of LIbra) on August 16, 2020
In this week's article, Palash Shah, the creator of Libra, co-authors an article about Libra with Sid Chadha
Libra AI is one of the most revolutionary tools for Neural Networks made in today's society of advancing artificial intelligence. Libra can easily create an optimized neural network in one line of code. Even with background knowledge in Machine Learning, its extremely cool to try this revolutionary software out. It can be applied to schools and colleges that would like to teach their students neural networks and how to create/analyze them. It works through taking into account a few factors, such as data and the amount of epochs you would like. It can then run a deep feed forward neural network, estimate the optimal layers size, and amount of layers to then return training accuracy, test accuracy, and loss over time. Afterwards, it can further analyze how good your model is with a confusion matrix and validation loss. It further returns a couple more fields if you choose to analyze it such as a ROC curve (an ROC curve is a graphical plot that illustrates the diagnostic ability of a binary classifier - similar to a positive/false positive chart).
How it Works?
Libra starts with a building block called queries, which acts as a way for users to interact with the functions and abilities of Libra. Libra uses and integrates technologies from all main Machine Learning frameworks. This includes, Tensorflow, a free and open source software library for dataflow and ML, Keras, another Machine Learning library, Sci-Kit learn, a regression and clustering Machine Learning library, and Huggingface, a text classifier or Natural Language processor that integrates Tensorflow. Natural language processing would be extremely useful to interpret datasets with text and/or audio, displaying Libra's versatility. Through using these integrations, Libra was made a possibility.
How to Install and Use it?
From Libra's GitHub, we see how to install it below:
Installation of Libra; GitHub
From Libra's GitHub, we see how to use it below Iit works in libra):
Usage of Libra; GitHub
Another, clear example of how to use Libra is also shown below
Libra data imporation in Jupyter (a Python IDE); YouTube
Libra data imporation in Jupyter (a Python IDE); YouTube
Analysis of Libra's model; YouTube
Overall, the application of Libra is huge with businesses and students all over the world creating and understanding more about neural networks in just a line of code.