Qsvm

algorithms/qml/qsvm/qsvm_bloch_sphere_feature_map.qmod

@@ -0,0 +1,16 @@
+qfunc bloch_feature_map_expanded___0(data: real[2], qba: qbit[1]) {
+  repeat (i: 1) {
+    RX(data[2 * i] / 2, qba[i]);
+  }
+  repeat (i: 1) {
+    RZ(data[(2 * i) + 1], qba[i]);
+  }
+}
+
+qfunc main(data1: real[2], data2: real[2], output qba: qnum<1, False, 0>) {
+  allocate(1, qba);
+  bloch_feature_map_expanded___0(data1, qba);
+  invert {
+    bloch_feature_map_expanded___0(data2, qba);
+  }
+}

algorithms/qml/qsvm/qsvm.synthesis_options.json → algorithms/qml/qsvm/qsvm_bloch_sphere_feature_map.synthesis_options.json

@@ -1,41 +1,42 @@
 {
   "constraints": {
+    "max_gate_count": {},
     "optimization_parameter": "no_opt"
   },
   "preferences": {
     "custom_hardware_settings": {
       "basis_gates": [
-        "t",
+        "u1",
+        "u2",
+        "z",
+        "rx",
         "cz",
-        "sx",
+        "t",
+        "u",
+        "id",
         "rz",
-        "rx",
+        "cx",
+        "tdg",
+        "sdg",
         "p",
-        "z",
-        "sxdg",
+        "ry",
         "h",
+        "y",
+        "sxdg",
+        "sx",
         "cy",
-        "u",
-        "r",
-        "u1",
-        "sdg",
-        "s",
-        "u2",
-        "tdg",
-        "id",
-        "ry",
         "x",
-        "cx",
-        "y"
+        "r",
+        "s"
       ],
       "is_symmetric_connectivity": true
     },
     "debug_mode": true,
     "machine_precision": 8,
-    "optimization_level": 3,
+    "optimization_level": 1,
     "output_format": ["qasm"],
     "pretty_qasm": true,
-    "random_seed": 1598682726,
+    "random_seed": 2336470090,
     "synthesize_all_separately": false,
     "timeout_seconds": 300,
     "transpilation_option": "auto optimize"

algorithms/qml/qsvm_pauli_feature_map/qsvm_pauli_feature_map.synthesis_options.json → algorithms/qml/qsvm/qsvm_pauli_feature_map.synthesis_options.json

@@ -1,32 +1,33 @@
 {
   "constraints": {
+    "max_gate_count": {},
     "optimization_parameter": "no_opt"
   },
   "preferences": {
     "custom_hardware_settings": {
       "basis_gates": [
-        "sdg",
-        "cx",
+        "u1",
+        "u2",
+        "z",
+        "rx",
+        "cz",
         "t",
-        "ry",
-        "rz",
         "u",
+        "id",
+        "rz",
+        "cx",
         "tdg",
-        "y",
+        "sdg",
+        "p",
+        "ry",
         "h",
-        "u1",
+        "y",
+        "sxdg",
         "sx",
-        "s",
-        "cz",
-        "x",
-        "p",
         "cy",
-        "rx",
-        "sxdg",
-        "id",
+        "x",
         "r",
-        "u2",
-        "z"
+        "s"
       ],
       "is_symmetric_connectivity": true
     },
@@ -35,7 +36,7 @@
     "optimization_level": 1,
     "output_format": ["qasm"],
     "pretty_qasm": true,
-    "random_seed": 3113768414,
+    "random_seed": 3912696462,
     "synthesize_all_separately": false,
     "timeout_seconds": 300,
     "transpilation_option": "auto optimize"

tests/notebooks/test_qsvm.py

@@ -6,20 +6,33 @@
 from testbook.client import TestbookNotebookClient
 
 
-@wrap_testbook("qsvm", timeout_seconds=204)
+@wrap_testbook("qsvm", timeout_seconds=300)
 def test_notebook(tb: TestbookNotebookClient) -> None:
     """
     A notebook for a hybrid classical quantum neural network.
     The test verifies that the pre-trained model is indeed well trained.
     """
     # test models
     validate_quantum_model(tb.ref("QSVM_BLOCH_SHPERE_qmod"))
+    validate_quantum_model(tb.ref("QSVM_PAULI_Z_ZZ"))
     # test quantum programs
     validate_quantum_program_size(
-        tb.ref_pydantic("qprog"),
+        tb.ref_pydantic("qprog_bloch"),
         expected_width=1,  # actual width: 1
         expected_depth=10,  # actual depth: 4
     )
-
+    validate_quantum_program_size(
+        tb.ref_pydantic("qprog_pauli"),
+        expected_width=2,  # actual width: 2
+        expected_depth=30,  # actual depth: 30
+    )
     # test notebook content
     assert tb.ref("test_score") == 1
+
+    # test notebook content
+    assert tb.ref("test_score_pauli") == 1
+
+    success_rate = tb.ref(
+        "100 * np.count_nonzero(predicted_labels_pauli == predict_real_labels_2) / len(predicted_labels_pauli)"
+    )
+    assert success_rate == 100