Posts

100DaysOfBlockchain: #3 - Where does Machine Learning come into play?

avatar of @aicoding
25
@aicoding
Ā·
0 views
Ā·
2 min read

I am guessing that if you are on Hive platform, you probably know all the basics of Blockchain technology and therefore the previous two posts might have been a bit boring to you.

Today I wanted to start brainstorming a bit about the posibilities of using Machine Learning for Blockchain technology and I am not necessarily talking about Bitcoin's price prediction. If I would know how to do so, I would probably not be here writing this post but drinking some Coconut Daikiri in Aruba šŸ¹

ā€‹ First of all, let's all get into the same page. Machine Learning is the combination of probability and statistics, linear algebra and computer science. In case you haven't heard... Machine Learning is the future. There are definitely a lot of exciting things that you can do with Machine Learning - BTW let's call it ML from now on. ML can help detect diseases, it can help you save money from fraudulent transactions and it can even help you pick your next purchase at Amazon.

ML learns from data, which means that is a self-adaptive software that it is constantly learning from new information. So far, I've only been looking at the very basics of the Blockchain technology - so it is not surprising that it is very difficult to myself, at this point, to think on how to improve it with other technology on top, such as ML algorithms. One thing that I do know is that Blockchain and Cryptocurrencies are tied together with lots of data. We all know that where is lots of data there is room for ML algorithms.

I found myself reading a research paper describing LearningChain, which is a blockchain-based machine learning framework which basically allows you to perform "decentralized" distributed machine learning (?). One of the main issues that machine learning faces every day are concerned with security and privacy of the data that is being used to feed these algorithms (you might think: obviously), it is actually not that obvious as it sounds. However, I do see potential on this idea, as the first idea that I would actually try to dive in into. Think about the possibilities of building machine learning models, where the privacy and security aspect of the data involved would not be an actual issue for the deployment and inference of the model.

It already feels overwhelming, I know - and there is just so much more I need to learn about this topic to even start thinking about developing anything of my own.

What are your thoughts here? Did you find this information useful?