verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
sumedhghaisas_ has quit [Ping timeout: 260 seconds]
sgupta has joined #mlpack
sumedhghaisas has joined #mlpack
sumedhghaisas has quit [Client Quit]
< ironstark>
rcurtin: I think its because of anaconda python distribution I am using. Will have to make certain changes accordingly. Trying to figure out what to do
< kris1>
Quick question why does the evlauate function in ffn require the arma::mat parameters as a parameter it is not being used anywhere
< zoq>
kris1: The optimizer expects that each function that should be optimized implements a specific set of functions with specific parameter the Gradient function is one of them.
< kris1>
Oh that’s why we have to have the same signature as well
< kris1>
Okay i got it thanks
< zoq>
Here to help :)
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
< kris1>
Is cross entropy function implmented in layers folder
< kris1>
i can’t seem to find it
< kris1>
I mean the binary cross entropy x.log(x’) + 1-x log (1 - x’)
< kris1>
where x’ is the predicted value
< lozhnikov>
kris1: no, I didn't see that in the layers folder
< kris1>
Okay so i think i would have to implement it i guess. I need it for the cd-k algorithm
< kris1>
for the evaluation criteria
< kris1>
I have implemented the pseudo likelihood but i would need to implement the cross entroy function
kris1 has quit [Quit: kris1]
< rcurtin>
kris1: I'd appreciate that if you did, I have some use for cross entropy... it would be nice if you could do more than binary too (like categorical cross-entropy)
sumedhghaisas_ has joined #mlpack
mikeling has joined #mlpack
sumedhghaisas_ has quit [Ping timeout: 260 seconds]
chenzhe_ has quit [Remote host closed the connection]
chenzhe has joined #mlpack
chenzhe has quit [Remote host closed the connection]
< sgupta>
rcurtin: hi! I have a registry running on master blaster already at port 5000
< sgupta>
rcurtin: I just copied it from my summer schedule, hence it is 15, 16 and so on. Sorry about that.
< sgupta>
rcurtin: where can I find tar ball of all versions of armadillo?
< rcurtin>
take a look in /opt/armadillo/, there are some there on masterblaster
< rcurtin>
other ones you will have to find from, e.g., srpms on the internet or the armadillo download page
< rcurtin>
sometimes Conrad likes to remove old versions from the download page, which can make it hard to find the old versions...
< rcurtin>
would you like to write a systemd service unit file for the docker registry container? this way we can know that the container starts on boot
< rcurtin>
it also might be useful to mount some directory to the registry container, so it can be stopped and restarted as needed and it won't lose the images
keonkim has joined #mlpack
< rcurtin>
I'll go ahead and get the docker environments set up to point to the registry container you're running, hang on
mentekid has quit [Quit: Leaving.]
< sgupta>
rcurtin: okay
< rcurtin>
I'll need to restart docker though, so is it ok if your registry container goes down?
< sgupta>
rcurtin: yes. I'll incorporate those changes and run it again
< rcurtin>
okay, let me know when the container goes down and I'll restart docker
< rcurtin>
I've also installed docker on dealgood and pointed its insecure registry at masterblaster
< rcurtin>
I also used iptables to block any access to masterblaster:5000:
< rcurtin>
iptables -A INPUT -p tcp -m tcp --dport 5000 -i eth0 -j DROP
< rcurtin>
dealgood can still access the registry though, since it has to tunnel port 5000 through ssh (which I set up)
< rcurtin>
those are all automatic on boot, so you shouldn't ever have to worry about them not being set up
< sgupta>
rcurtin: okay.
< sgupta>
rcurtin: I think you can restart. The container will stop automatically.
< sgupta>
rcurtin: I will map it to /home/sgupta/registry directory.
< rcurtin>
let's make that /home/jenkins/registry
< rcurtin>
I'll create the directory
< rcurtin>
docker is now restarted with the right --insecure-registry flag, let me know if there are any issues
< rcurtin>
zoq: gekko is back online, no data loss :)
sumedhghaisas_ has quit [Ping timeout: 260 seconds]
< zoq>
rcurtin: nice
kris1 has joined #mlpack
aashay has quit [Quit: Connection closed for inactivity]
govg has quit [Quit: leaving]
govg has joined #mlpack
kris1 has quit [Quit: kris1]
mikeling has quit [Quit: Connection closed for inactivity]
mentekid has joined #mlpack
aashay has joined #mlpack
ShangtongZhang has joined #mlpack
< ShangtongZhang>
Hi Marcus, I'm going to attend the DeepLearning and ReinforcementLearning summer school in Montreal during June 25 and July 6, in that time I may not able to contribute regularly,
< ShangtongZhang>
hope my previous work before June can make up for that time
< zoq>
ShangtongZhang: I think if you continue with the same pace as you do, we will be ahead the schedule by June 25.
< zoq>
ShangtongZhang: Btw. really nice work and nice to see you here :)
< ShangtongZhang>
^.^
sumedhghaisas has joined #mlpack
< rcurtin>
ShangtongZhang: sounds like a fun time :)
< rcurtin>
NIPS was in Montreal for a few years, fun city to travel to
< ShangtongZhang>
Yeah, and it's my first time to Montreal
ShangtongZhang has quit [Quit: Page closed]
< zoq>
I agree Montreal is really nice, I can recommend Mount Royal, great hike and great view