Re: More memory-efficient inner product for large last
- To: mathgroup at smc.vnet.net
- Subject: [mg106880] Re: [mg106870] More memory-efficient inner product for large last
- From: "Vincent N. Virgilio" <virgilio at ieee.org>
- Date: Tue, 26 Jan 2010 06:33:01 -0500 (EST)
- References: <201001251009.FAA09421@smc.vnet.net>
On Mon, Jan 25, 2010 at 9:16 AM, Leonid Shifrin <lshifr at gmail.com> wrote: > Hi Vince, > > I suggest that you use lazy matrix multiplication, which can be implemented > for example as follows: > > > SNIP > As can be seen, my version is less memory-efficient for list-to-list dot > product, but vastly > more efficient for other operations. I did not test on such huge lists as > your original ones since > I don't have so much memory at my disposal at the moment (running Eclipse > and SQLDeveloper), > but I would expect similar effect. > > Hope this helps. > > Regards, > Leonid > > Leonid, Phenomenal work! Yours saved ~ 1.5GB RAM over mine (for 1.2M elements). Thank you very much. Vince Virgilio
- References:
- More memory-efficient inner product for large last dimension?
- From: Vince Virgilio <blueschi@gmail.com>
- More memory-efficient inner product for large last dimension?