commit

natural_images/.gitignore

 *.pt
+*.npy

wf_variants/main.py

+import tikzplotlib
 import wf_variants
 import torch
 import numpy as np
@@ -7,6 +8,22 @@ from pprint import pprint
 import time
 from datetime import timedelta
 start = time.time()
+def tikzplotlib_fix_ncols(obj):
+    """
+    workaround for matplotlib 3.6 renamed legend's _ncol to _ncols, which breaks tikzplotlib
+    """
+    if hasattr(obj, "_ncols"):
+        obj._ncol = obj._ncols
+    for child in obj.get_children():
+        tikzplotlib_fix_ncols(child)
+# def tikzplotlib_fix_ncols(obj):
+#     """
+#     workaround for matplotlib 3.6 renamed legend's _ncol to _ncols, which breaks tikzplotlib
+#     """
+#     if hasattr(obj, "_ncols"):
+#         obj._ncol = obj._ncols
+#     for child in obj.get_children():
+#         tikzplotlib_fix_ncols(child)
 # Python using Pytorch version of https://github.com/hubevan/reshaped-Wirtinger-flow/blob/master/exampleRWF.m
 # Example of the RWF and IRWF algorithm under 1D Gaussian designs
 # The code below is adapted from implementation of the TWF desinged by Y. Chen and E. Candes, Wirtinger Flow algorithm designed and implemented by E. Candes, X. Li, and M. Soltanolkotabi
@@ -23,17 +40,23 @@ rwf_err = wf_variants.rwf(A,x,y)
 irwf_err = wf_variants.irwf(A,x,y,batch= 1)
 imrwf_err = wf_variants.irwf(A,x,y,batch= 64)
 
-plt.semilogy(np.arange(0, np.size(wf_err)), wf_err, '-r',label="Wirtinger Flow")
-plt.semilogy(np.arange(0, np.size(twf_err)), twf_err, '-m',label="Truncated Wirtinger Flow")
+plt.semilogy(np.arange(0, np.size(wf_err)), wf_err, '-r',label="WF")
+plt.semilogy(np.arange(0, np.size(twf_err)), twf_err, '-m',label="TWF")
 # plt.semilogy(np.arange(0, np.size(itwf_err)), itwf_err, '-b')
-plt.semilogy(np.arange(0, np.size(rwf_err)), rwf_err, '-c',label="Reshaped Wirtinger Flow")
-plt.semilogy(np.arange(0, np.size(irwf_err)), irwf_err, '-g',label="Incrementally Reshaped Wirtinger Flow")
-plt.semilogy(np.arange(0, np.size(imrwf_err)), imrwf_err, '-y',label="Incrementally Minibatched Reshaped Wirtinger Flow")
+plt.semilogy(np.arange(0, np.size(rwf_err)), rwf_err, '-c',label="RWF")
+plt.semilogy(np.arange(0, np.size(irwf_err)), irwf_err, '-g',label="IRWF")
+plt.semilogy(np.arange(0, np.size(imrwf_err)), imrwf_err, '-y',label="IMRWF")
 plt.xlabel('Iteration')
 plt.ylabel('Relative error (log10)')
 plt.title('Wirtinger Flow Variants')
 plt.legend()
-plt.savefig('Wirtinger Flow Variants.pdf')
+# fig = plt.figure()
+fig = plt.gcf()
+tikzplotlib_fix_ncols(fig)
+tikzplotlib.save("wf_variants.tex")
+
+# plt.savefig('Wirtinger Flow Variants.pdf')
+# tikzplotlib.save("wf_variants.tex")
 print("misson accomplished in:")
 end = time.time()
 elapsed_time = end - start

wf_variants/wf_variants.py

@@ -13,7 +13,7 @@ params['n1']          = 64  # signal dimension
 params['m']           = 10 * params['n1']  # number of measurements
 params['cplx_flag']   = 1  # real: cplx_flag = 0;  complex: cplx_flag = 1;
 params['grad_type']   = 'TWF_Poiss'  # 'TWF_Poiss': Poisson likelihood
-params['T']           = 1500  # number of iterations
+params['T']           = 800  # number of iterations
 params['npower_iter'] = 30  # number of power iterations
 npower_iter = params['npower_iter']  # Number of power iterations
 tau0 = 330