--- Log opened Wed Aug 09 00:00:44 2017
00:10 -!- augur [~augur@hostelingintl.static.monkeybrains.net] has joined ##hplusroadmap
00:14 -!- augur [~augur@hostelingintl.static.monkeybrains.net] has quit [Ping timeout: 260 seconds]
00:29 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
00:35 -!- augur is now known as 17SABNV0E
00:38 -!- 17SABNV0E [~augur@162.245.20.162] has quit [Ping timeout: 268 seconds]
00:40 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
00:45 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 268 seconds]
00:53 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
00:58 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 260 seconds]
01:10 < archels_> .title https://www.youtube.com/watch?v=guI-Uw4W0T8
01:10 < yoleaux> Computational Design of Telescoping Structures (SIGGRAPH 2017) - YouTube
01:47 -!- Netsplit *.net <-> *.split quits: jrayhawk, axlshear, soylentbomb, justan0theruser, superkuh, HEx1
01:50 -!- augur [~augur@hostelingintl.static.monkeybrains.net] has joined ##hplusroadmap
01:52 -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-xsyqjhdmnsdqmmvj] has joined ##hplusroadmap
01:54 -!- augur [~augur@hostelingintl.static.monkeybrains.net] has quit [Ping timeout: 240 seconds]
02:00 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
02:01 -!- CRM114 [~urchin@unaffiliated/urchin] has joined ##hplusroadmap
02:04 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 255 seconds]
02:31 -!- Netsplit over, joins: jrayhawk, soylentbomb
02:31 -!- Netsplit over, joins: justan0theruser, axlshear
02:33 -!- superkuh [~superkuh@unaffiliated/superkuh] has joined ##hplusroadmap
02:34 -!- HEx1 [~HEx@hexwab.plus.com] has joined ##hplusroadmap
03:00 -!- bluebear_ [~dluhos@80.95.97.194] has joined ##hplusroadmap
03:08 -!- jaboja [~jaboja@jaboja.pl] has joined ##hplusroadmap
04:13 < pasky> kanzure: yeah that's neat; I'd sum it up as "series of non-linear random projections of the same kind NNs use preserves essential information [for solving various popular tasks]"; you'd probably also do decent imagenet on the final image representation, but i'd be curious how deep a network would you need for that
04:15 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 276 seconds]
04:18 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
04:24 < kanzure> hm.
04:25 < kanzure> i've been trying to find where in tensorflow i had to do a modified logarithm or whatever, for using keras_backend.ctc_batch_cost
04:25 < kanzure> but nothing is showing up in my notes.
04:28 < kanzure> this line was problematic: y_pred = tf.log(tf.transpose(y_pred, perm=[1, 0, 2]) + 1e-8)
04:31 < kanzure> oh, nevermind.
04:47 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
04:51 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 260 seconds]
05:16 -!- preview [~quassel@2407:7000:842d:4026:49b:aa53:f0af:a741] has joined ##hplusroadmap
05:17 -!- jaboja [~jaboja@jaboja.pl] has quit [Ping timeout: 276 seconds]
05:36 -!- CRM114 is now known as nukular
05:46 -!- preview [~quassel@2407:7000:842d:4026:49b:aa53:f0af:a741] has quit [Read error: No route to host]
05:47 -!- mihajul [4e619c89@gateway/web/freenode/ip.78.97.156.137] has joined ##hplusroadmap
05:47 -!- preview [~quassel@2407:7000:842d:4026:49b:aa53:f0af:a741] has joined ##hplusroadmap
05:50 < mihajul> http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)31442-3/fulltext
05:52 < mihajul> hi.. is there a paperbot around here ?
05:54 < cluckj> nope
05:56 < mihajul> ok :( thanks anyway
05:57 -!- mihajul [4e619c89@gateway/web/freenode/ip.78.97.156.137] has quit [Quit: Page closed]
06:14 < kanzure> he ded
07:02 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
07:07 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 240 seconds]
07:07 < wrldpcmbl> Try #icanhazpdf with link to paper on twitter
07:09 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
07:17 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 260 seconds]
07:28 -!- fltrz [d5d38b31@gateway/web/freenode/ip.213.211.139.49] has quit [Ping timeout: 260 seconds]
07:29 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
07:36 -!- fltrz [d5d38b31@gateway/web/freenode/ip.213.211.139.49] has joined ##hplusroadmap
07:40 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 260 seconds]
07:57 < fltrz> the "random weight" network refers to https://github.com/mileyan/random_weights but it dont exist?
08:04 < fltrz> this is a slightly newer version of the paper, but it misses the github link above https://arxiv.org/pdf/1606.04801v2.pdf
08:09 < kanzure> the pruning paper was pretty neat, getting a 500kb network.  i wonder if they looked at what the network actually looks like.
08:09 < kanzure> for a lot of this feature detection stuff, i'm convinced it can be transformed into a list of simple declarative classification rules
08:10 < kanzure> simple, interpretable models (like "x + 1") is way better than "here's a 5 MB blob of black magic").
08:11 -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-xsyqjhdmnsdqmmvj] has quit [Quit: Connection closed for inactivity]
08:12 < TMA> fltrz: mileyan pushed to random_weights at mileyan/neural-style. ;; https://github.com/mileyan/neural-style/tree/random_weights
08:12 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 240 seconds]
08:14 < kanzure> "add random weight feature" https://github.com/mileyan/neural-style/commit/e85500d180a16737e559a0f5d898966106d5ef3e
08:15 < fltrz> requirements: Pre-trained VGG network ?
08:18 < fltrz> I don't see any weights actually being randomized here https://github.com/mileyan/neural-style/commit/e85500d180a16737e559a0f5d898966106d5ef3e
08:18 < kanzure> yes it looks like "random == True" was previously implemented
08:19 < fltrz> oh
08:20 < fltrz> to me it seems like random aperture coding
08:21 < kanzure> seems like the weights were from a caffee model dunno how random that is https://github.com/mileyan/neural-style/blob/random_weights/reverse_rw.py
08:22 < fltrz> presumably he did not want to reinitialize the whole model with random weights on each execution, and stored his random weights in a model once
08:31 -!- nukular [~urchin@unaffiliated/urchin] has quit [Quit: That's it for today]
08:57 -!- Burnin8 [~Burn@pool-96-241-130-178.washdc.fios.verizon.net] has joined ##hplusroadmap
09:00 -!- Burn_ [~Burn@pool-96-241-130-178.washdc.fios.verizon.net] has quit [Ping timeout: 260 seconds]
09:23 < maaku> the good old define(random) = 4 // chosen by dice roll
09:27 -!- andytoshi [~apoelstra@wpsoftware.net] has quit [Changing host]
09:27 -!- andytoshi [~apoelstra@unaffiliated/andytoshi] has joined ##hplusroadmap
10:14 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
10:17 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Client Quit]
10:17 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
10:21 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 240 seconds]
10:22 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
10:22 < nmz787> https://github.com/JuliusKunze/speechless
10:22 < nmz787> from the mozilla deepspeech IRC room
10:25 < kanzure> "Generating adversarial examples for speech recognition" http://web.stanford.edu/class/cs224s/reports/Dan_Iter.pdf
10:28 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
10:44 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
10:55 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has joined ##hplusroadmap
11:07 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Read error: Connection reset by peer]
11:08 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
11:08 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has joined ##hplusroadmap
11:11 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
11:19 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
11:34 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has quit [Quit: Malvolio]
11:40 -!- fltrz [d5d38b31@gateway/web/freenode/ip.213.211.139.49] has quit [Ping timeout: 260 seconds]
12:07 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 240 seconds]
12:09 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
12:15 < JayDugger> What has Hans Moravec done since 2003?
12:16 < kanzure> he was replaced by hod lipson
12:18 < JayDugger> Okay, point taken. Specifically, Moravec?
12:20 -!- bluebear_ [~dluhos@80.95.97.194] has quit [Quit: Leaving.]
12:20 < kanzure> looks like he's been writing patents https://scholar.google.com/scholar?q=author%3A%22hans+moravec%22&hl=en&as_sdt=0%2C44&as_ylo=2003&as_yhi=
12:20 -!- Guest77357 is now known as abetusk
12:21 < JayDugger> Yeah, saw that, but thank you.
12:44 < kanzure> "STARDATA: A StarCraft AI Research Dataset" https://arxiv.org/abs/1708.02139 https://github.com/TorchCraft/StarData
12:58 -!- aeiousomething [~aeiousome@183.82.170.54] has quit [Ping timeout: 240 seconds]
13:33 < kanzure> cool: "All this madness is managed through the MAINTAINERS file and the get_maintainers.pl script, which for any given snippet of code can tell you who’s the maintainer, who should review this, where the right git repo is, which mailing lists to use and how and where to report bugs. And it’s not just strictly based on file locations, it also catches code patterns to make sure that ...
13:33 < kanzure> ...cross-subsystem topics like device-tree handling, or the kobject hierarchy are handled by the right experts."
13:33 < kanzure> from http://blog.ffwll.ch/2017/08/github-why-cant-host-the-kernel.html
13:37 < kanzure> "The way it’s reviewed is usually not the way it’s merged, instead one of the subsystems is selected as the leading one and takes the pull requests, as long as all other maintainers agree to that merge path. Usually it’s the subsystem most affected by a set of changes, but sometimes also the one that already has some other work in-flight which conflicts with the pull request. Sometimes ...
13:37 < kanzure> ...also an entirely new repository and maintainer crew is created, this often happens for functionality which spans the entire tree and isn’t neatly contained to a few files and directories in one place. A recent example is the DMA mapping tree, which tries to consolidate work that thus far has been spread across drivers, platform maintainers and architecture support groups."
13:44 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
13:44 -!- helleshin [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has joined ##hplusroadmap
13:59 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
14:21 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has joined ##hplusroadmap
14:42 -!- Qfwfq [Qfwfq@gateway/shell/suchznc/x-cqgdbkoelttxmgcy] has quit [Quit: SuchZNC - Such ZNC, many free, w0w... -- https://suchznc.net]
16:15 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has quit [Read error: Connection reset by peer]
16:19 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has joined ##hplusroadmap
16:19 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 240 seconds]
16:40 -!- Qfwfq [Qfwfq@gateway/shell/suchznc/x-ngzxhrzkzsvvfznj] has joined ##hplusroadmap
16:40 -!- Qfwfq [Qfwfq@gateway/shell/suchznc/x-ngzxhrzkzsvvfznj] has quit [Changing host]
16:40 -!- Qfwfq [Qfwfq@unaffiliated/washirving] has joined ##hplusroadmap
16:40 -!- Qfwfq [Qfwfq@unaffiliated/washirving] has quit [Changing host]
16:40 -!- Qfwfq [Qfwfq@gateway/shell/suchznc/x-ngzxhrzkzsvvfznj] has joined ##hplusroadmap
16:45 -!- atrus6 [~atrus6@72.241.82.247] has joined ##hplusroadmap
17:00 < kanzure> hmph
17:59 -!- preview [~quassel@2407:7000:842d:4026:49b:aa53:f0af:a741] has quit [Remote host closed the connection]
18:02 -!- preview [~quassel@2407:7000:842d:4000::3] has joined ##hplusroadmap
18:03 -!- mrdata [~mrdata@unaffiliated/mrdata] has joined ##hplusroadmap
18:03 -!- danfox [~danfox@vulpinedesigns.co.uk] has quit [Ping timeout: 276 seconds]
18:04 -!- danfox [~danfox@vulpinedesigns.co.uk] has joined ##hplusroadmap
18:14 < kanzure> i would be okay with models taking 10x longer to create/train as long as they are substantially more interpretable/understandable/less black box spaghetti magic.
18:27 -!- fltrz [d5d38b31@gateway/web/freenode/ip.213.211.139.49] has joined ##hplusroadmap
18:28 < fltrz> from the squeezenet paper: "Strategy 1. Replace 3x3 filters with 1x1 filters." what is a 1x1 filter? isn't that just a single weight?
18:36 -!- mrdata [~mrdata@unaffiliated/mrdata] has quit [Ping timeout: 240 seconds]
18:37 -!- mrdata [~mrdata@unaffiliated/mrdata] has joined ##hplusroadmap
19:11 -!- UnknownRogue [~UnknownRo@184.75.213.38] has joined ##hplusroadmap
19:14 -!- mrdata [~mrdata@unaffiliated/mrdata] has quit [Read error: Connection reset by peer]
19:43 -!- yashgaroth [~yashgarot@2606:6000:cd4d:3300:9d6f:c41c:f922:6906] has joined ##hplusroadmap
20:05 -!- UnknownRogue [~UnknownRo@184.75.213.38] has quit [Read error: Connection reset by peer]
20:45 < kanzure> hmmmmph
20:49 < kanzure> cool the docker people are finally fixing their build system https://github.com/moby/moby/issues/20304#issuecomment-321433741
20:50 < kanzure> "Optimal brain damage" http://yann.lecun.com/exdb/publis/pdf/lecun-90b.pdf
21:11 -!- Cory [~Cory@unaffiliated/cory] has quit [Ping timeout: 240 seconds]
21:19 -!- Cory [~Cory@unaffiliated/cory] has joined ##hplusroadmap
21:54 -!- helleshin [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has quit [Ping timeout: 240 seconds]
23:09 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has quit [Remote host closed the connection]
23:12 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has joined ##hplusroadmap
23:12 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has quit [Remote host closed the connection]
23:18 -!- yashgaroth [~yashgarot@2606:6000:cd4d:3300:9d6f:c41c:f922:6906] has quit [Quit: Leaving]
23:51 -!- augur [~augur@162.245.20.162] has joined ##hplusroadmap
23:55 -!- augur [~augur@162.245.20.162] has quit [Ping timeout: 240 seconds]
--- Log closed Thu Aug 10 00:00:45 2017