Good info in this artice I believe.
http://ncr101.montana.edu/Light1994C...ula%20text.htm

All of it is interesting, but specifically scroll down to the "Pulsed Lighting" section. Basically, applied light was pulsed at about 6.7kHz, with duty cycles of 1%, 5%, 10%, and the constant 100%. They were all done with the same time-integrated photon flux (ie. the 1% had 100x higher intensity than the 100% one).

Result: for tomato plants, they all gave about the same photosynthetic response. So to match a constant-on LED source, you'd need to be able to drive the peak output 1/(duty cycle) times higher. Seems like one would be better off forgetting about pulsing, and overdriving their LED array as much as their thermal dissipation allows.

thoughts:
*might be different for cannabis vs. tomato leaves
*might get different results pulsing a broader range of wavelengths
*duty cycles might have greater effect if pulsed at lower frequencies. I didn't find anything about constant duty cycle over a range of frequencies, aside from:

http://www.freepatentsonline.com/5012609.html
"It has been experimentally demonstrated that when light flashes of 100 microseconds duration follow each other too quickly, the light is not fully utilized (B. Kok. 1956. Photosynthesis in flashing light. Biochim. et Biophys. Acta, 21:245-258.). Extending the time of the dark period to about 20 milliseconds increased the relative "yield" per flash of light."

I don't know why they used 100us pulses, but here they're suggesting about 0.5% duty cycle at 50Hz as being optimal, with no mention of light intensity affecting photosynthesis. If it's true that the plant sees this case as the same as DC (when both have equal peak outputs), then this would give lots of room to overdrive the LEDs for great gains. But if it still requires equal integrated outputs, then it's not really worth it (100x intensity requires much more than 100x current since relationship gets non-linear at currents ~>100mA, with baseline around 20mA). This is probably something that needs to be found out to get the most out of LEDs.