Testing Some Crazy New Math For Neural Network Tuning!

[@RafaelZanini-km4so]

If we take 8HP transmission as example, there is always interpolation between shift ID, road resistance, ... It is hard to believe that engine will not interpolate automatically tables in case "CAM angle is in between". Nevertheless, great video! I totally appreciate your work!!!

[@GoatRopeGarage]

It isn’t so much that there isn’t interpolation, it’s that the tables simply don’t exist, they are created from formulas, the possible number of tables are endless. Is this approach necessary? Probably not but I love testing new approaches. This is just one way I’m testing tuning the neural networks!

[@claytonargo3085]

Super glad you’re doing nn! We learn from our failures more than we learn from our successes. I think your parabolic math makes good sense on the face of it but only testing will tell. I like where you’re going with that! Someday I want to put a stage 1 cam in my 2014 rt. Comp cams says lockout vvt or use a cam phaser 7 degrees limit. With only 7 degrees cam swing would that develop only 2 nn tables or just 1? (Vs 5 cam angle tables with stock cam/vvt, 50 degrees swing) Super bonus dumb ass (me) question: if you have any third gen hemi and your cam can swing 50 degrees without hitting pistons is it big enough?

[@teagreen2220]

I think the real solution is aftermarket ECUs.

[@alexjohnson5434]

Locking the cams out and cleaining up the 5 tables individually doesn't seem all that terrible, for sure not optimal though. Now doing that on a DOHC Pentastar where the cams swing into like 17 of the 25 tables, that would be a nightmare. One thing that might be worth a try would be using a .CSV export and doing the math in excel rather than the HP Tuners software.

[@GoatRopeGarage]

The current math abilities in the scanner wouldn’t be effective for a DOHC, but I think you’re right that it could be done in excel or even MegaLog

[@maalmull]

I started to hate these things more NN Sucks!