Conversation
|
Sorry for the late response on this one. I had no prior HTML/CSS/JS experience so it is taking me time to learn them and understand the codes. I am currently watching a series of youtube lectures on D3.js and learning how to build webpages using Java Script. It shall take me around 1-2 more days. I know that it is after GSoC period ends but I will continue to contribute and complete the webpage. In the meantime I have worked on benchmarking the Shark C++ ML library and will open up a PR on it by tomorrow. |
|
Sure, no worries. If you have any comments on this view it would be great, if not that's ok too. :) |
|
Ok, here is the actual display I was aiming for: Use the 'All metric plots for parameter sweeps'. The database there is only results from logistic regression runs. I think, before I merge this, I will fix some of the logistic regression code and add/modify some datasets (not all of the ones there give interesting results for logistic regression). Basically, the display plots a learning curve for parameter sweeps over all libraries for a given method. This means that if we benchmark different optimizers they will always show up on the same plot. |
|
This looks great to me, also really like the updated tooltip. Not really related to this PR, but do you think we should update the library color scheme, I think it's sometimes difficult to distinguish the different libs. |
|
Sure, I would agree with changing the colors. Let me look into the colormap we are using and see if I can come up with something better. |
This is not the prettiest JS ever written, but it adds two views to plot both runtimes over a sweep, and metrics vs. runtime over a sweep. Here is an image of each view: