spare gcca

Structure of Project

The structure of this project should be contruct like this:

• your_project_name/
  • gcca/
    • spare_gcca
    • origin_gcca
    • ...
  • gcca_data/
    • csv_data/
    • genes_data/
    • twitter_data/
    • three_view_data/
    • weight/synthetic_data_model_list_U.pickle
    • image/plot_for_spare

Package

You should install some necessary package within python3

keras
theano
numpy
sklearn
pandas

Test result

Main programe is in all_kinds_of_test.py. You should open this file and choose which model you want to run

Choose whatever you want and have fun:

Which model do you want to choose?
0.gcca,
1.spare_gcca,
2.cca,
3.deepcca,
4.WeightedGCCA,
5.dgcca_
>>>0
Which result do you want to test?
0.test in gene data,
1.test in gene data for std,
2.test in gene data whether normalize or not,
4. nothing
>>>0

And also, you can check the file which you like and deepen it.

parameter tunning

In deep cca, you only can tune epoch , batch size and learning rate, because the other parameter do not make a big deal. Or maybe you can edit the code whatever you want:

In deep_cca.py:

class deepcca(metric):
    def __init__(self, ds, m_rank, batch_size = 50, epoch_num = 10, learning_rate = 1e-3):
        
		# ...

        # parameter you can tune
        self.batch_size = batch_size
        self.epoch_num = epoch_num
        self.learning_rate = learning_rate

In dgcca_format.py, you only can tune epoch and batch size:

class dgcca_(metric):
    def __init__(self, ds, m_rank, batchSize=40, epochs = 200):
        
        # ...

        # parameter you can tune
        self.batchSize = batchSize
        self.epochs = epochs

Feedback

If you have any issue, please let me know or email me [email protected] or [email protected]

Reference

Thanks to the code of wgcca written by abenton and deep cca written by VahidooX