I keep thinking it might be useful for forecasting and pattern recognition in training data but by the time I get home it's the last thing I want to do.
Power meters · Public discussion
Anyone tried using data mining and neural net software on CP data?
Thread details
What we know about this thread
- Original section
- Power meters
- Published
- 12 January 2006
- Last activity
- 15 January 2006
- Original author
- mises
- Posts
- 5
- Discussion status
- Public discussion
- Total views
- 1,567
- Views / 30 days
- 0
The navigation and discussion metadata provide context. Posts remain in their original chronological order.
-
-
mises said:
I keep thinking it might be useful for forecasting and pattern recognition in training data but by the time I get home it's the last thing I want to do.
You mean something like this?
-
acoggan said:
You mean something like this?
Exactly! Thanks for the link. I guess I will have to get motivated now. No freaking way I would use neural nets in the stock market but since the causal factors are presumably knowable and quantifiable for training and performance I thought it might be a good application. I have no problem using black box methods if it saves me from having to use my brain for an activity I like predominantly for its physical dimensions. -
Would such modeling be more valuable if the cyclist was riding against a precise power script and if the script was known (available in data form)?
-
RapDaddyo said:
Would such modeling be more valuable if the cyclist was riding against a precise power script and if the script was known (available in data form)?
To use the relationships a neural net derives to construct a training plan it's much simpler if the inputs are very well defined.A fairly decent free program for data mining and neural nets can be gotten at:
http://www.sciengy.net/rpf/The help files explain the main ways of using it fairly well. As long as you have a reason to presume some causal link it doesn't hurt to throw everything you can think of into the mix (morning bp, hr, weight, body temp, psychological measures, etc) and weed it out after seeing what it comes up with after training the net. It's also much better if you have enough data so that training and testing the results can be done on 2 different data sets to avoid curve fitting.
Active in the last 60 minutes
Active in this thread
0 users · 0 guests ·0 bots ·0 total
No signed-in users are active right now.
No known search crawlers active right now.