Where is the line between fitting and actual patterns? - page 64

[TheXpert]

fxsaber:

That's the way it is.

A pig finds dirt everywhere. If you want to fit, you can fit.

If you want a truthful result and have no illusions, roll forward rules, and even just forward is passable.

[fxsaber]

Комбинатор:

A pig finds dirt everywhere. If you want to fit, you can fit.

If you want a truthful result and have no illusions, roll forward rules, and even just forward is passable.

It's an eternal topic of debate.

[fxsaber]

Комбинатор:

If you want a true result and there are no illusions, roll forward rules, and even just forward is passable.

There is some terminological confusion. In my opinion, when they say "forward", they mean two situations

1. TS is optimized with forward. I.e. forward is being run on each pass of the optimizer.
2. Optimization is running in full. Only single runs of the already completed optimization are run on the forward - OOS.

1 - self-deception. 2 is a normal practice.

[SidorOFF]

Комбинатор:

A pig finds dirt everywhere. If you want to fit, you can fit.

If you want a truthful result and have no illusions, roll forward rules, and even just forward is passable.

The result will be a fitting forward.

[Yury Kirillov]

SidorOFF:
The result will be a fit for the forward.

+