semi-supervised nonnegative matrix factorization with convex coefficients
- Python 3.x
- cvxopt >= 1.2.5
- numpy >= 1.19.5
- pandas >= 1.2.0
- scipy >=1.6.0
- six >=1.15.0
| V | Path to .tsv or .csv file for V matrix. Assumes no header exists. Matrix of M features (e.g. CpG sites) over N raw samples. |
| -W | Path to .tsv or .csv file for W matrix. Can be partial. Assumes no header exists. Atlas matrix of M features over K bases (e.g. tissues). |
| -H | Path to .tsv or .csv file for H matrix. Assumes no header exists. Coefficients matrix of K bases over N samples such that each column takes the convex combination of a single sample. |
| -lb, --lower_bound | Float representing the lower bound for the values of the matrices V and W. Valid values are greater or equal to zero. Default: 0 |
| -ub, --upper_bound | Float representing the upper bound for the values of the matrices V and W. Valid values are greater than zero. Default: 1 |
| -c, --free_w_cols | Number of free columns to add to W. Default: 0 |
| -iw, --init_w | Comma separated string, stating the type of distribution (first argument) and parameters (second and so on) for W initialization. default="normal,0,1". Valid distribution values: normal,beta. |
| -ih, --init_h | Comma separated string, stating the type of distribution (first argument) and parameters (second and so on) for H initialization. default="beta,70,100". Valid distribution values: normal,beta. |
| -t, --iter_num | Number of iteration for the algorithm. Default: 10 |
| -r, --reps | Number of Repetitions for the algorithm. Default: 5 |
| -o, --tol | Tolerance Parameter. Default: 1e-5 |
python ssnmfcc.py V.tsv -c 3
(1.) python ssnmfcc.py V.tsv -W W.tsv
(2.) python ssnmfcc.py V.tsv -W W.tsv -c 2
python ssnmfcc.py V.tsv -H H.tsv -r 4 --init_w "beta,30,70"