Great, now it gets interesting. 
The model output is where the price should be and thus can be used as indicator/forecast. Is that correct? I take a deeper look into the docs in a few minutes.
I'm not into NN, but what I was asking myself for quite a while: can it be used to predict the future and how accurate is that? In the charts the output is pegged to the actual value, but can it produce an output for the future derived from the past? And if so, could it be used transfered to something else, i.e. weather forecast?
Last one: we know the future weather is based on the current and past weather and can be determined. Can we say for sure markets are predictable?
Sorry for turning this in an neural networks AMA.
First one, correct. Model output is the market-implied price based upon regression of input values throughout historical samples during each network's training. The primary indicator provided is a combination of diverg and dSdev, which provide both direction and entry/exit indicators for trading the target instrument. Additionally, in some strategies, output price may be considered as a target exit point once a position has been taken.The model output is where the price should be and thus can be used as indicator/forecast. Is that correct? I take a deeper look into the docs in a few minutes.
I'm not into NN, but what I was asking myself for quite a while: can it be used to predict the future and how accurate is that? In the charts the output is pegged to the actual value, but can it produce an output for the future derived from the past? And if so, could it be used transfered to something else, i.e. weather forecast?
Last one: we know the future weather is based on the current and past weather and can be determined. Can we say for sure markets are predictable?
Sorry for turning this in an neural networks AMA.
Second one, yes, to some degree. There are plenty of modeling applications already in use which attempt to do this very same thing. Although it is my own personal experience that these platforms and strategies often fail during periods of high volatility, which is consequently the very conditions where our platform excels. Weather is a considerably more difficult task to manage in any context, as the dimensionality of data is extreme. Where we may process data on up to 30 or so inputs on our most complex models, weather modeling may handle many thousands on relatively simplistic models. They utilize some of the largest supercomputers in the world because they really have no choice.
It cannot be said that markets are always predictable. Indeed, the future cannot be predicted by any man or machine. However, it can be said that market events follow statistical patterns, providing those with a suitable understanding of said patterns with a demonstrable advantage. I am not among those who understand these patterns, so instead, created the next best thing: a machine learning platform which creates specialist models who do.
