You need to define what output you want for a given input. You train the neural network by supplying both the input and the output, and it makes the internal adjustments to produce those results.
In the examples, they provide 0 and 1 as input, and expect 1 as output, and 1 and 1 as input, and expect 0 as output. That's what an XOR does.
You need to define a set of inputs and the output you expect in each case. That is what you will use to train your neural network.
| [reply] |
Hi,
Thanks.
For example, NASDAQ Index "XYZ", the following values are the daily rate .
"2010-02-03" = 23.25;
"2010-02-04" = 32.02;
"2010-02-05" = 28.21;
"2010-02-06" = 12.25;
Can you suggest me what output I can define for the above input data .
how do I predict the value for "2010-02-07" ?
Please correct me if I am missing anything
Raja K
| [reply]
[d/l] |
The common approach is to train your network on the data you have, and then use the last data item as the goal.
In your example, you would train your network on the days for 2010203, 2010204, 2010205, and then test it (for example) against 2010206.
Soon after, you will encounter the miracle of Overfitting your model to your training data.
Also, you will find that if (say) neural networks were good at predicting "the market", everybody would already be using them.
| [reply] |