naywhayare 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/
< paul>
<p>where <img class="formulaInl" alt="$ \beta $" src="doxygen/form_45.png"/> is the vector of regression coefficients.</p> <p>If <img class="formulaInl" alt="$ \lambda_1 > 0 $" src="form_46.png"/> and <img class="formulaInl" alt="$ \lambda_2 = 0 $" src="doxygen/form_47.png"/>, the problem is the LASSO. If <img class="formulaInl" alt="$ \lambda_1 > 0 $" src="form_46.png"/> and <img class="formulaInl" alt="$ \lambda_2 > 0 $" src="form_48.
< paul>
some png miss the /doxygen
< paul>
therefore they can't be display properly
paul has quit [Quit: Page closed]
Anand has joined #mlpack
Anand has quit [Ping timeout: 246 seconds]
Anand has joined #mlpack
< Anand>
Marcus : Had a look at the code?
< marcus_zoq>
Anand: Yeah!
< marcus_zoq>
Anand: We need to add the new bootstrap task to the config parser file.
< Anand>
Which file?
< marcus_zoq>
Anadn: Ah, no, I've already modified the parser so that we can add any task. Sorry for the confusion!
< Anand>
Marcus : I was also thinking that will it be better to take the maximum rather than average while normalizing (we divide by 100 to normalize as of now)?
< marcus_zoq>
Anand: Hmm, I guess this shouldn't be a huge difference.
< Anand>
Yeah, the average will always be less
< Anand>
The difference can be large sometimes. But, yes may be average presents a better picture
< Anand>
Btw, is 100 good enough or should we increase that too?
Anand has quit [Ping timeout: 246 seconds]
< marcus_zoq>
Anand: The guys from the paper use 1000 iterations. But for now let's use 100 iterations. Afterwards we can check if should increase the iterations. Does this make sense?
Anand has joined #mlpack
andrewmw94 has joined #mlpack
Anand has quit [Ping timeout: 246 seconds]
Anand has joined #mlpack
Anand has quit [Ping timeout: 246 seconds]
Anand has joined #mlpack
< Anand>
Marcus : What I meant was, once we have the metrics dictionary (after normalization), it is only for a single library, right?
< Anand>
Earlier I was in the impression that we have metrics for a single method for all libraries and not metrics for all methods of a single library.
< Anand>
Also we have 3 things : 1.Various metrics for a method, 2.The method itself and 3.The library. How can we display all the 3 in our table?
< marcus_zoq>
Anand: Actually we iterate through all methods (line: 145) and for a method we iterate through all libraries (line: 171). So the easiest thing you can do is to instantiate a dictionary in line 166. So we can use the dict to store the results. We can print a single table for each method and use the library name as column index and the metric name as row index. Does this sound reasonable?
< Anand>
Ok! So, our parser parses the config file taking a method for all libraries?
< Anand>
And then taking the next method?
< marcus_zoq>
Anand: Right!
< Anand>
Alright, sorry for the confusion.
< marcus_zoq>
Anand: So everything is already sorted.
< Anand>
So, we will have a table for each method then.
< Anand>
I think so
< marcus_zoq>
Anand: Unless you have a better idea?
< Anand>
No, actually the reference paper is not dealing with multiple libraries. But we are.
< Anand>
I think what we are thinking now should be good
< Anand>
Marcus : Can you have a look now? I have also added a print function
< marcus_zoq>
Anand: Sure!
< marcus_zoq>
Anand: 'NameError: name 'datasets' is not defined' -> Line 296
< marcus_zoq>
Anand: You should use dataset.
< Anand>
Marcus : line 212, for dataset in datsets: We are using "datsets" instead of "datasets". Just a spelling mistake!
udit_s has joined #mlpack
< marcus_zoq>
Anand: Can you run the bootstrap task?
< Anand>
Marcus : No, debugging!
< marcus_zoq>
Anand: Maybe this helps -> 'NameError: name 'metric_dict_key' is not defined'
< Anand>
Marcus : Have a look now, There were quite a few bugs. It works now