 |
Angelo Aquino
2015-04-01 09:10:39
|
|
Hi Good day. I just want to ask two questions:
For the Neural Network
1. How I can DECREASE the error, network and MSE?
2. How can I increase the number of decimal places used by the Testing stage? It appears that only 3 decimal places are used but I need FOUR decimal places. How can this be changed??
Many thanks in advance.
|
|
|
 |
QuantShare
2015-04-01 12:57:11
0
|
|
Hi,
1/ Network errpr and MSE are calculated automatically during training and they are the result of your network, inputs and other settings. You cannot control these directly.
2/ Which field are you referring to?
|
|
|
|
|
Angelo Aquino
2015-04-01 13:05:20
0
|
|
Hi thanks for the quick response. For the decimal place question I am referring to the Test values in the Neural network training window: the predicted values vs the actual values. I input data up to 5 decimal places but in the window showing the predicted values vs the actual values only 3 decimal places is used, which is not an acceptable order for the error.. Should be 4-5 decimal places.
How do I change it? Many thanks again.
|
|
|
|
|
QuantShare
2015-04-02 02:41:38
0
|
|
Hi,
The table is showing only 3 decimal places but the internal calculation uses all available decimal places.
|
|
|
|
|
Angelo Aquino
2015-04-02 06:07:51
0
|
|
Thank you.
Okay so as I understand it, Quantshare cannot display the decimal places beyond 3 decimals in the Predicted vs Actual values?
If this is a setting we cant change, the next logical question would be How do I improve my Neural Network? What settings should I use to lessen the errors? Because , related to the last question, I have to reduce the error up to the 4th decimal place. A nonzero error in the third decimal place is not acceptable..
Many thanks again.
|
|
|
|
|
QuantShare
2015-04-03 02:30:35
0
|
|
As said previously, the calculation uses all decimal places, however the control just displays three. This is only for the display.
We will make an update to increase the number of decimal displayed.
|
|
|
|
|
Angelo Aquino
2015-04-03 03:10:57
0
|
|
Thank you very much.
I mean I want to ask about how to make the Neural Networks more accurate ?
|
|
|
|
|
QuantShare
2015-04-03 05:58:47
0
|
|
Not sure I understand what you mean by "make the Neural Networks more accurate"
|
|
|
|
|
Angelo Aquino
2015-04-04 07:29:08
0
|
|
By a more accurate Neural Network, I mean decreased absolute errors.
What settings or adjustments can I do to the network to effectively decrease the absolute error?
Thank you.
|
|
|
|
|
QuantShare
2015-04-06 02:02:44
0
|
|
The absolute error is the result of the predicated values vs actual values. This depends on your network type, settings and most importantly inputs.
|
|
|
|
|
Torsten
2016-03-05 05:11:32
0
|
|
Hi QS,
as I always run some sample checks in excel to ensure that my AMM scripts are running correctly, I also would appreciate more decimal places for MMPosition.EntryPrice and MMPosition.LastPrice.
It seems, that QS is calculating with all 5 digits, but the output of the variables is limited to 4 digits.
Best regards,
Torsten
|
|
|
|
|
0
