Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Chris Hokamp
    @chrishokamp
    hi everyone
    post your questions or problems about the code
    and we'll try to help
    Chris Hokamp
    @chrishokamp
    working on some ipython notebooks here
    Chris Hokamp
    @chrishokamp
    note that def compute_scores in hmm.py is a little confusing
    the emission and transition matrices get copied for each time step
    David Buchaca Prats
    @davidbp
    Screen Shot 2015-07-03 at 17.16.20.png
    It is confusing but the code is adapted to work with this. We can change it though if we all agree.
    I would vote for a change in this case. It doesn't make any sense to store the same matrix again and again...
    also understand what happens with the feature mapper when you initialize the class
    notice that it delegates to discriminative_sequence_classifier.py
    which then delegates to sequence_classifier
    we need to call build_features on a feature_mapper in order to initialize all of the feature templates https://github.com/LxMLS/lxmls-toolkit/blob/student/lxmls/sequences/id_feature.py#L44-L55
    reinhack
    @reinhack

    Does anyone know how to force the confusion matrix plot to always have the same Y-scale? For exercise 3.1-3.2 I can't really see better results with the extended feature because the confusion matrix plot is plotted in different scale.

    This is the code I used to produce the confusion matrix plot:

    import lxmls.sequences.confusion_matrix as cm
    import matplotlib.pyplot as plt
    confusion_matrix = cm.build_confusion_matrix(test_seq.seq_list, pred_test, len(corpus.tag_dict), crf_online.get_num_states())
    cm.plot_confusion_bar_graph(confusion_matrix, corpus.tag_dict, xrange(crf_online.get_num_states()), 'Confusion matrix')
    plt.show()

    Chris Hokamp
    @chrishokamp
    checking that now
    Maria Karanasou
    @mkaranasou
    @reinhack: do you mean something like: "plt.axis([min(x), max(x), min(y), max(y)])" adjusted accordingly?
    Daniel Ferreira
    @dcferreira
    @reinhack you can probably put something like fig.set_ylim([min, max]) in the plot_confusion_matrix function, and I guess it should work
    just confirmed that it works
    Chris Hokamp
    @chrishokamp
    very cool sorry i got distracted with questions
    Daniel Ferreira
    @dcferreira
    correction: it's not the plot_confusion_matrix function, it's plot_confusion_bar_graph
    reinhack
    @reinhack
    Cheers guys, fig.set_ylim() did the trick.