RoninAi Updates: Ai Neural Networks Advance [Pushing Boundaries]

As promised in our last Ai updates article (which you can read here), we are back for another round of exciting updates in the RoninAi development process. Progress has been excellent up to this point and we are very excited to begin to solidify a more concrete time frame for our big launch.

In this update, we will follow through on the selection of pre-configured neural networks and user settings for those networks, as well as the data-based autotuning module that we received a lot of questions about following the release of our last Ai update. We will also go over one very important upcoming milestone in development that will be reached in the coming days that we are very excited to showcase.

Curation of pre-configured neural networks

As discussed in the previous Ai update, RoninAi has the power to be the optimal tool for every investor, including those who manage millions of dollars with a dedicated team of experts. However, for the individual investor who may not be experienced with harnessing technology as powerful as this, offering pre-configured neural networks goes a long way to help make RoninAi the best piece of software it can possibly be for traders of all levels of experience and expertise.



Selection of user settings for pre-configured neural networks

Configuring the user settings for the pre-configured neural networks was a task that the development team knew went hand in hand with curating the pre-configured networks themselves. The ability for every type and style of user to leverage the incredible analysis that RoninAi does through artificial intelligence and machine learning is core to what the RoninAi development team believes in.

Source data-based autotuning modele

The autotuning module of RoninAi is one of the most technically advanced and fascinating parts of RoninAi. Getting to see it in its finished state is a great indicator of both present and future continued success of the software. Since the Ai is constantly being fed market data, it is able to not only analyze the current data through its black box to give an accurate prediction of future market states, it is able to feed this data through this autotuning module that will ever so slightly adjust the ways in which the Ai analyzes markets. This allows RoninAi to get even stronger and even more precise as time goes on, allowing those who use it to leverage its power to feel its positive effects even more clearly over time.



After this very successful and exciting period where many important parts of the final RoninAi product were completed, the development team now, in the short term, turns its attention squarely towards one singular feature of the final product.

Implementation of the LSTM functionality

While other Ai updates had a look at several different tasks that the development team was undertaking, this one will have just one, as the team is currently laser-focused on getting it done right. LSTM or Long Short-Term Memory, stores data in a unique way, allowing for context to influence output. To simplify, if a non-LSTM recurrent neural network (RNN) is given data of cryptocurrency prices over periods of time, it will likely take a simplistic view of patterns and trends. However, an LSTM model has the ability to separate short-term memory and store it, prioritize certain pieces of information, and be able to put both historical and current data into context to achieve a more precise output.

The memory cells learn when to allow data to enter, leave or be deleted through the iterative process of making guesses, back-propagating error, and adjusting weights via gradient descent.

The diagram below illustrates how data flows through a memory cell and is controlled by its gates.

For example, an RNN model might remember that markets trend downwards in Q1, but an LSTM model can put other vital information into context and more accurately predict just how much, if at all, it will go up or down based on all available factors.

A simple machine learning model or an Artificial Neural Network may learn to predict the cryptocurrency prices based on a number of features: volume, value etc. While the price of the cryptocurrency depends on these features, it is also largely dependent on numerous factors and historical data. In fact for a trader, these values in the previous days (or the trend) is one major deciding factor for predictions.

In the conventional feed-forward neural networks, all test cases are considered to be independent. That is when fitting the model for a particular day, there is no consideration for the cryptocurrency prices (trends) on the previous days.


This is also true for analyzing social sentiment, fork analysis, blockchain protocol, etc. How heavily has the news affected crypto prices? Or when a cryptocurrency forks? LSTM takes these past occurrences into consideration when making predictions and proves invaluable to traders.

With LSTM functionality in play, RoninAi will begin to truly resemble its impending finished product, as it is one of the core pillars of what makes RoninAi so special.

In following updates, we will have more updates about RoninAi’s exciting development as the software begins to reach and resemble its completed product. Stay tuned to as we near launch date, and don’t forget to get your pre-sale discount on both the Investor and Pro Terminals of RoninAi here:

Related article: RoninAi Updates: A Look Inside Cutting Edge Ai Developments