MANTA Test

4 min readNov 4, 2021

8–12 November

In accordance with the Stage 2 of our roadmap and following on from the tests of MANAS and MANIA in October, we are now opening up MANTA (Matrix AI Network Training Algorithm) for testing.

MANTA means a lot to MATRIX because of the following facts:
1. The computing power of un-selected nodes, instead of idling, can be used to train an algorithm for benefits of the miners.
2. Lower the barrier for people to develop AI services or applications, which will bring more AI services to MANAS.
3. Enriches the varieties of Matrix’s ecosystem.
4. Increases the liquidity of MAN.

As the name MANTA suggests, it is about training an AI algorithm. Unlike MANAS and MANIA, which can be tested by general users, which most of them had fun with trying to play around or fool the systems, MANTA is more for people with AI background.

This time, we will control to limited person because of the server deployment. If you are interested and would like to access the test please send an email to Eric Choy with brief description about your background in AI.

We had demo videos about MANTA back in July (Link 1, Link 2). The reason we chose demonstration videos to present MANTA is because of the same reason as above, i.e. it is more for AI scientists or people with AI background to train their models, and hard to present it to the general public. That said, after the hard work of the project team, we move forward from video to a platform ready for test.

The differences between the two version are as follows:
1. Stage 1, is about distributing tasks to multi-node and work simultaneously. It has not fulfilled the auto machine learning part.
2. Stage 2, the platform is more ready with front-end and back-end, plus task progress monitoring.
3. Stage 2, it has two distributed auto training modules, namely, Image Classification and Depth Estimation.

This progress is important in the sense that
1. This is a ready-to-use platform for AI scientists.
2. Auto Machine Learning becomes possible, and computing power and tasks become meaningful with it.

Now that the background to this test is clear and its importance emphasised allow us to go further with the test scope and details.

A. Objective:

To demonstrate to the public the technology progress we have achieved in distributed auto machine learning.
To solicit comments from people with AI knowledge on the operation and workflow.

B. Details

From 8 to 12 November
Open to people with AI knowledge and testing to be done in allocated time slots
Interested persons, please send an email to EricChoy@Matrix.io mentioning your AI background.
A test guide will be provided.
For those who would like to know more about MANTA, we will also shoot a video going through the process.

I know you are curious about MANTA so we have added some screenshots below for you to enjoy.