Temporal replacement of algo with SVM

bob/pad/face/config/pytorch/autoencoder/mc_pad_config_bw_nir_d_128x128_face_10_relu.py

@@ -80,35 +80,55 @@ auto-encoder.
 #=======================================================================================
 # define algorithm:
 
-from bob.pad.face.algorithm import CustomFunc
-import torch.nn.functional as F
-
-
-CONFIG_FILE = "autoencoder/face_features_net_128x128/batl/batl_mlp_algorithm_celeba_and_partial_3_layers_batl_features/batl_mlp_10_relu_algorithm.py"
-CONFIG_GROUP = "bob.pad.face.config.pytorch"
-MODEL_FILE = "/idiap/temp/onikisins/project/ODIN/experiment_data/pad_experiments_bob4/batl_db/autoencoder_face_features_128x128/batl_train_mlp_10_relu_algo_celeba_and_partial_3_layers_batl_features/experiment_8/model_17.pth"
-INVERT_SCORES_FLAG = True
-
-def _prediction_function(local_model, x):
-    x = x.squeeze()
-    x = local_model.fc1(x)
-    x = F.relu(x)
-    # second fully connected NO sigmoid:
-    x = local_model.fc2(x)
-    return x
-
-KWARGS = {}
-KWARGS["config_file"] = CONFIG_FILE
-KWARGS["config_group"] = CONFIG_GROUP
-KWARGS["model_file"] = MODEL_FILE
-KWARGS["invert_scores_flag"] = INVERT_SCORES_FLAG
-KWARGS["prediction_function"] = _prediction_function
-
-FUNCTION_NAME = "net_forward"
-
-algorithm = CustomFunc(function_name = FUNCTION_NAME,
-                       frame_level_scores_flag = True,
-                       kwargs = KWARGS)
+from bob.pad.base.algorithm import SVM
+
+MACHINE_TYPE = 'C_SVC'
+KERNEL_TYPE = 'RBF'
+N_SAMPLES = 10000
+TRAINER_GRID_SEARCH_PARAMS = {
+    'cost': [2**P for P in range(-3, 14, 2)],
+    'gamma': [2**P for P in range(-15, 0, 2)]
+}
+MEAN_STD_NORM_FLAG = True  # enable mean-std normalization
+FRAME_LEVEL_SCORES_FLAG = True  # one score per frame(!) in this case
+
+algorithm = SVM(
+    machine_type=MACHINE_TYPE,
+    kernel_type=KERNEL_TYPE,
+    n_samples=N_SAMPLES,
+    trainer_grid_search_params=TRAINER_GRID_SEARCH_PARAMS,
+    mean_std_norm_flag=MEAN_STD_NORM_FLAG,
+    frame_level_scores_flag=FRAME_LEVEL_SCORES_FLAG)
+
+#from bob.pad.face.algorithm import CustomFunc
+#import torch.nn.functional as F
+#
+#
+#CONFIG_FILE = "autoencoder/face_features_net_128x128/batl/batl_mlp_algorithm_celeba_and_partial_3_layers_batl_features/batl_mlp_10_relu_algorithm.py"
+#CONFIG_GROUP = "bob.pad.face.config.pytorch"
+#MODEL_FILE = "/idiap/temp/onikisins/project/ODIN/experiment_data/pad_experiments_bob4/batl_db/autoencoder_face_features_128x128/batl_train_mlp_10_relu_algo_celeba_and_partial_3_layers_batl_features/experiment_8/model_17.pth"
+#INVERT_SCORES_FLAG = True
+#
+#def _prediction_function(local_model, x):
+#    x = x.squeeze()
+#    x = local_model.fc1(x)
+#    x = F.relu(x)
+#    # second fully connected NO sigmoid:
+#    x = local_model.fc2(x)
+#    return x
+#
+#KWARGS = {}
+#KWARGS["config_file"] = CONFIG_FILE
+#KWARGS["config_group"] = CONFIG_GROUP
+#KWARGS["model_file"] = MODEL_FILE
+#KWARGS["invert_scores_flag"] = INVERT_SCORES_FLAG
+#KWARGS["prediction_function"] = _prediction_function
+#
+#FUNCTION_NAME = "net_forward"
+#
+#algorithm = CustomFunc(function_name = FUNCTION_NAME,
+#                       frame_level_scores_flag = True,
+#                       kwargs = KWARGS)
 
 """
 In this case, score is predicted by a pre-trained MLP, **excluding** output sigmoid.