What to feed to the input of the neural network? Your ideas... - page 73

 
It sounds like you've explored a lot of ways to analyze price action, but if you're hitting a wall, it might help to rethink your approach. You could try looking at momentum and volatility metrics like ATR or Bollinger Band width over N periods to measure shifts in volatility. Another idea is to compare the strength of EUR and USD across a broader range of pairs, like EUR/JPY or USD/JPY, to get a better feel for market momentum. You could also focus on market structure—like recent swing highs/lows or breakouts—to add context to your analysis. If you're into more advanced methods, using order flow proxies like volume or tick data could provide insights into market activity. Lastly, try considering time-based factors, like how certain patterns behave at specific times, like during session openings or after key news releases. If you're really into quant, feeding these features into a machine learning model might reveal hidden relationships.
 
elettra fischer machine learning model can reveal hidden correlations.

Thanks for the ideas

elettra fischermachine learning model can reveal hidden correlations.

Others are dealing with MO, they have their own branch.

The level is easier here.

 

N.S.A. needs weights, if I understand correctly.

That's what you need to submit.

 
Renat Akhtyamov #:

NSka needs weights, if I understand correctly.

That's what you need to submit.

Rena, you misunderstand. Some data is fed to the NS input, but not weights. And weights of NS synapses are determined by its type, architecture, training...
and have nothing to do with input data. Weights can even change in dynamics.

 
Grigori.S.B #:

Rena, you misunderstand. The input of the NS is some data, but not weights. And weights of NS synapses are determined by its type, architecture, training ...
and have nothing to do with input data. Weights can even change in dynamics.

Weights change dynamically? I haven't heard such fantasies yet
 
mytarmailS #:
Do weights change over time? That's a fantasy I've never heard before
What are weights?
 
Ivan Butko #:


Playing with parameters using creative methods (mixing formulas by poking and prodding) can sometimes improve results, again - conditionally.

In the example below we managed to increase the number of forward trades again, the chart started to look better, but at the end it started to fade. + The backtest outside the optimisation period (from 2000 to 2012) became a victim, it failed.

Forward 2021-2025 for EUR






I still have a feeling that the market is described by some kind of looped formula (some chart parameters should be connected with each other somehow) and such an owl will work almost everywhere. Thoughts aloud

Feed volumes to the neural network input
 
osmo1717 #:
What are weights?
Google it
 
mytarmailS #:
Do weights change over time? I've never heard that fantasy before.
What's so surprising? Changed sample --> changed model parameters. Moreover, even on the same sample the parameters may differ due to differences in parameter initialisation and as a consequence different convergence (the neural network will simply converge to a neighbouring local minimum, for example).
 
Evgeniy Chernish #:
What's so surprising? The sample has changed --> model parameters have changed. Moreover, even on the same sample the parameters may differ due to differences in parameter initialisation and as a consequence different convergence (the neural network will simply converge to a neighbouring local minimum, for example).
When the neural network is trained, its weights are unchanged. Only the input data change