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/
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
kris1 has quit [Client Quit]
kris_ has quit []
kris_ has joined #mlpack
kris1 has joined #mlpack
mikeling has joined #mlpack
govg has quit [Ping timeout: 240 seconds]
sumedhghaisas_ has joined #mlpack
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
sumedhghaisas_ has quit [Ping timeout: 240 seconds]
sumedhghaisas_ has joined #mlpack
sgupta has quit [Ping timeout: 240 seconds]
sgupta has joined #mlpack
vivekp has joined #mlpack
govg has joined #mlpack
vivekp has quit [Ping timeout: 268 seconds]
vivekp has joined #mlpack
sumedhghaisas_ has quit [Ping timeout: 240 seconds]
mentekid has quit [Quit: Leaving.]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid1 has joined #mlpack
mentekid has quit [Read error: Connection reset by peer]
mentekid1 has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
shikhar has joined #mlpack
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has quit [Quit: kris1]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has joined #mlpack
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has quit [Quit: kris1]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid1 has joined #mlpack
mentekid1 has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has joined #mlpack
mentekid has joined #mlpack
mentekid has quit [Client Quit]
vivekp has quit [Ping timeout: 276 seconds]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
vivekp has joined #mlpack
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Read error: Connection reset by peer]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
chenzhe has quit [Ping timeout: 240 seconds]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Read error: Connection reset by peer]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has quit [Quit: kris1]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has joined #mlpack
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
mentekid has joined #mlpack
mentekid has quit [Client Quit]
kris1 has quit [Quit: kris1]
< rcurtin>
I think Yannis' client has some connection issues... :)
govg has quit [Ping timeout: 268 seconds]
govg has joined #mlpack
aashay has quit [Quit: Connection closed for inactivity]
champion has joined #mlpack
champion has quit [Client Quit]
aashay has joined #mlpack
mentekid has joined #mlpack
mikeling has quit [Quit: Connection closed for inactivity]
< sgupta>
rcurtin: hi! I am facing a lot of errors installing GCC from source. Tried a lot of blogs, the official website too.
< sgupta>
rcurtin: I kind of need help with this.
chenzhe has joined #mlpack
kris1 has joined #mlpack
aashay has quit [Quit: Connection closed for inactivity]
vivekp has quit [Ping timeout: 258 seconds]
chenzhe has quit [Quit: chenzhe]
vivekp has joined #mlpack
vivekp has quit [Ping timeout: 268 seconds]
vivekp has joined #mlpack
shikhar has quit [Quit: WeeChat 1.7]
< rcurtin>
sgupta: sure, I am here to help, it is probably easier if you send an email with the full documentation of what you've done and the output that you've received
kris1 has quit [Quit: kris1]
vivekp has quit [Ping timeout: 240 seconds]
kris1 has joined #mlpack
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
aashay has joined #mlpack
sumedhghaisas_ has joined #mlpack
< sumedhghaisas_>
zoq: Hey Marcus. I have completed the batch norm layer for convolutional layer. How should we test that? same as we did for feedforward network?
< sumedhghaisas_>
zoq: huh... interesting. Will have to check my code again for this. Will do that.
< sumedhghaisas_>
okay and for that comment you posted on the pull request
< sumedhghaisas_>
I am fixed everything except for the last one
< zoq>
yeah, interesting issue
< zoq>
okay, great :)
< sumedhghaisas_>
I am not sure I understand that right
< zoq>
just posted an updated
< sumedhghaisas_>
ahh okay. So we can bypass extra creation of zero vector
< sumedhghaisas_>
let me see if I can do that without much overhead
< sumedhghaisas_>
there must some way with standard library functions
< zoq>
If there's an easy way to do that, it might be an interesting micro optimization, if not, I guess it's not worth the hassle.
< sumedhghaisas_>
I think all the optimizations in NN are micro optimizations :)
< sumedhghaisas_>
ahh about the idea posted by Ryan on the SGD optimization calls. What do you think about that?
< sumedhghaisas_>
if we remove the Evaluate call all together... user may not be able to plot the training curves though
< sumedhghaisas_>
next week hopefully I will start the implementation of NTM... now that the RNN framework supports variable length I can test them without hassle. Really excited...
< zoq>
Nice, excited to start with NTM :)
< zoq>
Let me take a look at #1030.
< sumedhghaisas_>
I also took a look at #1005... the copy task implementation. How is the current LSTM and GRU doing there?
< sumedhghaisas_>
that can be the baseline for the NTM
< zoq>
hm, I think let's say there is an option useTolerance set to true, it's still a good idea to avoid the extra call of the Evaluation function.
< sumedhghaisas_>
ahh okay so we do both...
< zoq>
That's what I propose, yes.
< zoq>
About the Copy task. Both models are able to learn the standard copy task, (tested length >= 20, so we have to test it on length > 20 and see if we can find a length that's not "learnable".
< sumedhghaisas_>
Fine. Will do that.
< sumedhghaisas_>
I see. I checked his implementation though. He is changing Rho... that might class with the new implementation
< sumedhghaisas_>
I mean with variable length one
< sumedhghaisas_>
*clash
< zoq>
yeah, we have to updated that, once your code is merged.
< zoq>
Also, both LSTM and GRU failed on the Repeat Copy task ... I guess the way we incorporate the repeat information is suboptimal.
< sumedhghaisas_>
according to the paper they should work on Repeat Copy task... something might be wrong from our side
< sumedhghaisas_>
I had another minor suggestion in RNN framework if you have time...
< zoq>
Waiting for the Add and Sort task, if they work, I'll think about another way to integrate the repeat information.
< zoq>
sure, go ahead
< sumedhghaisas_>
So with the dummy Evaluate we are bypassing the 1 extra Evaluate call. But if you notice at the end of SGD... there is one another Evaluate call.
< zoq>
the one for the overalobjective?
< sumedhghaisas_>
With our current way of optimizing each training example separately we make 1 extra Evaluate call per training data
< sumedhghaisas_>
yup... for returning
< sumedhghaisas_>
I think that can be avoided too
< sumedhghaisas_>
if we create a Train function which accepts the field
< zoq>
yeah, definitely
benchmark has joined #mlpack
benchmark has quit [Client Quit]
< zoq>
if you like open a PR, happy to merge it :)
< sumedhghaisas_>
okay... :) I am running out of time so fast... haha
< zoq>
If you don't have time to do it, it's okay :)
< sumedhghaisas_>
also working on dissertation... the idea I discussed with you remember? According to my current testing it seems to be learning ...
< sumedhghaisas_>
don't know if it works faster that policy gradient descent... if it does... that would be super cool :)
< zoq>
oh, I agree, but that would be relatively easy to test?
< sumedhghaisas_>
I am currently thinking on that. There are couple of things. So I am using MCTS on top of the softmax network... so a match with my method takes longer than the match with pure policy gradient descent.
< sumedhghaisas_>
So comparing the performance with each training match might not be a good idea...
< zoq>
ah right I remember
< sumedhghaisas_>
on the other hand my implementation of MCTS in not an optimal one... on top of that its in python... so even comparing time is little tricky.
< rcurtin>
I'm marking slake offline, since I think both dewang and I are using it for prototyping/testing the benchmark system
< rcurtin>
that also means moving the PR builder for benchmarks to gekko (chosen arbitrarily)
< rcurtin>
if that's an issue let me know and I can pick a different one :)
< sumedhghaisas_>
zoq, rcurtin: The style checks are still failing for my PR. But I can't find any error... in the report. How do you detect the style errors?
< rcurtin>
sumedghaisas_: link to failure report? I can take a glance
< rcurtin>
looks like mostly "line ends in whitespace", I think technically we could automate the removal of that whitespace
< rcurtin>
but I'm not sure that's high enough on anyone's list :)
< rcurtin>
this syntax checker that abhinav helped put together is a huge time saver already (well... for the people reviewing PRs that is :))
< sumedhghaisas_>
ahh I see... I should stop using stupid GEdit now :)
< zoq>
sumedhghais: I see comparing time is difficult, so number of steps might be an option.
< zoq>
rcurtin: Sounds fine for me.
< sumedhghaisas_>
zoq: thats also an option... number of training example given to each network.
< rcurtin>
zoq: I wanted to get your opinion on a change to the benchmarking system
< rcurtin>
if I want to benchmark logistic regression, what people really care about is "how quickly does it converge to a good solution"
< rcurtin>
so really what I would want to provide is a learning curve with several points on it for each method, showing the runtime it took to get to each accuracy value
< rcurtin>
in order to do this, I would need some support for "sweeping" some parameter (in this particular case it would be max_iterations)
< rcurtin>
it seems to me, like I can add another table 'sweep_info' and then have each run reference a particular sweep_info_id and sweep_element_id, so I can then assemble a full "sweep" to plot the values
< rcurtin>
then in the configuration side, I could specify a sweep with ':', i.e. options: "-n 10:10:100" would sweep over [10, 20, 30, 40, ..., 100]
< rcurtin>
what do you think, is this a reasonable way to do it, or should I approach the problem differently?
< rcurtin>
need to go to bed, it is 1am here... I'll be back tomorrow :)
< zoq>
rcurtin: hm, yeah I guess that's a reasonable approach, we could also add parameter range support e.g. in the config file, the main benchmark script runs the method on the range and appends the results in the current table, I guess your solution is probably easier to parse since you can work with SQL.