Start the python port of demo_fact_hadamard.m into pyfaust.demo.hadamardfact class.

At this stage the hadamard transform and its full matrix are properly generated through FaustFactory, then the factorization operates with a non-significant relative error (so it works).

wrapper/python/pyfaust/demo.py

@@ -453,4 +453,28 @@ class runtimecmp:
         #tight_layout()
         #show()
 
+class hadamardfact:
+
+    def run():
+        from pyfaust import FaustFactory
+        from pyfaust.factparams import ParamsHierarchicalFact, ConstraintInt, \
+        ConstraintName, StoppingCriterion
+        n = 5
+        d = 2**n
+        H = FaustFactory.hadamard(n)
+        full_H = H.toarray()
+
+        params = ParamsHierarchicalFact(n, is_update_way_R2L=True, init_lambda=1.0,
+                                        fact_constraints=[ConstraintInt(ConstraintName(ConstraintName.SPLINCOL),d,d,2)
+                                        for i in range(0,n-1)],
+                                        res_constraints=[ConstraintInt(ConstraintName(ConstraintName.SPLINCOL),d,d,int(d/2.**(i+1)))
+                                         for i in range(0,n-1)],
+                                        data_num_rows=d, data_num_cols=d,
+                                        stop_crits=[StoppingCriterion(num_its=30),StoppingCriterion(num_its=30)],
+                                        is_verbose=False)
+        had_faust = FaustFactory.fact_hierarchical(H.toarray(), params)
+        full_had_faust = had_faust.toarray()
+        rel_err = norm(full_had_faust-full_H)/norm(full_H)
+        print("\n\nRelative error between hadamard matrix and its transform: ",
+              rel_err)