change model - add dense before server output in new model

add some new run scripts

main.py

@@ -156,6 +156,7 @@ def main_train(param=None):
     logger.info("compile and train model")
     embedding.summary()
     model.summary()
+    logger.info(model.get_config())
     model.compile(optimizer='adam',
                   loss='binary_crossentropy',
                   metrics=['accuracy'] + custom_metrics)

models/pauls_networks.py

@@ -68,10 +68,10 @@ def get_new_model(dropout, flow_features, domain_features, window_size, domain_l
     ipt_domains = Input(shape=(window_size, domain_length), name="ipt_domains")
     ipt_flows = Input(shape=(window_size, flow_features), name="ipt_flows")
     encoded = TimeDistributed(cnn)(ipt_domains)
-
+    merged = keras.layers.concatenate([encoded, ipt_flows], -1)
-    y2 = Dense(1, activation="sigmoid", name="server")(encoded)
+    y = Dense(dense_dim, activation="relu")(merged)
-    merged = keras.layers.concatenate([encoded, ipt_flows, y2], -1)
+    y2 = Dense(1, activation="sigmoid", name="server")(y)
-
+    # CNN processing a small slides of flow windows
     y = Conv1D(cnn_dims,
                kernel_size,
                activation='relu',

models/renes_networks.py

@@ -51,14 +51,16 @@ def get_new_model(dropout, flow_features, domain_features, window_size, domain_l
     ipt_domains = Input(shape=(window_size, domain_length), name="ipt_domains")
     ipt_flows = Input(shape=(window_size, flow_features), name="ipt_flows")
     encoded = TimeDistributed(cnn)(ipt_domains)
-
+    merged = keras.layers.concatenate([encoded, ipt_flows], -1)
-    y2 = Dense(1, activation="sigmoid", name="server")(encoded)
+    y = Dense(dense_dim, activation="relu")(merged)
-    merged = keras.layers.concatenate([encoded, ipt_flows, y2], -1)
+    y2 = Dense(1, activation="sigmoid", name="server")(y)
-
+    # CNN processing a small slides of flow windows
-    y = Conv1D(cnn_dims,
+    y = Conv1D(filters=cnn_dims, kernel_size=kernel_size, activation='relu', padding="same",
-               kernel_size,
+               input_shape=(window_size, domain_features + flow_features))(y)
-               activation='relu',
+    y = MaxPool1D(pool_size=3, strides=1)(y)
-               input_shape=(window_size, domain_features + flow_features))(merged)
+    y = Conv1D(filters=cnn_dims, kernel_size=kernel_size, activation='relu', padding="same")(y)
+    y = MaxPool1D(pool_size=3, strides=1)(y)
+    y = Conv1D(filters=cnn_dims, kernel_size=kernel_size, activation='relu', padding="same")(y)
     # remove temporal dimension by global max pooling
     y = GlobalMaxPooling1D()(y)
     y = Dropout(dropout)(y)

run.sh

@@ -4,30 +4,29 @@ python main.py --mode train \
                 --train /tmp/rk/currentData.csv \
                 --model /tmp/rk/results/simple_both \
                 --epochs 25 \
-                --hidden_char_dims 128 \
+                --embd 64 \
+                --hidden_chaar_dims 128 \
                 --domain_embd 32 \
                 --batch 256 \
                 --balanced_weights \
                 --model_output both
 
-python main.py --mode test --batch 512 --model /tmp/rk/results/simple_both --test /tmp/rk/futureData.csv --model_output both
-
 python main.py --mode train \
                 --train /tmp/rk/currentData.csv \
                 --model /tmp/rk/results/simple_client \
                 --epochs 25 \
+                --embd 64 \
                 --hidden_char_dims 128 \
                 --domain_embd 32 \
                 --batch 256 \
                 --balanced_weights \
                 --model_output client
 
-python main.py --mode test --batch 512 --model /tmp/rk/results/simple_client --test /tmp/rk/futureData.csv --model_output client
-
 python main.py --mode train \
                 --train /tmp/rk/currentData.csv \
                 --model /tmp/rk/results/simple_new_both \
                 --epochs 25 \
+                --embd 64 \
                 --hidden_char_dims 128 \
                 --domain_embd 32 \
                 --batch 256 \
@@ -35,17 +34,65 @@ python main.py --mode train \
                 --model_output both \
                 --new_model
 
-python main.py --mode test --batch 512 --model /tmp/rk/results/simple_new_both --test /tmp/rk/futureData.csv --model_output both
-
 python main.py --mode train \
                 --train /tmp/rk/currentData.csv \
                 --model /tmp/rk/results/simple_new_client \
                 --epochs 25 \
+                --embd 64 \
                 --hidden_char_dims 128 \
                 --domain_embd 32 \
                 --batch 256 \
                 --balanced_weights \
                 --model_output client \
                 --new_model
+##
+##
+python main.py --mode train \
+                --train /tmp/rk/currentData.csv \
+                --model /tmp/rk/results/simple_both \
+                --epochs 25 \
+                --embd 64 \
+                --hidden_chaar_dims 128 \
+                --domain_embd 32 \
+                --batch 256 \
+                --balanced_weights \
+                --model_output both \
+                --type rene
 
-python main.py --mode test --batch 512 --model /tmp/rk/results/simple_new_client --test /tmp/rk/futureData.csv --model_output client
+python main.py --mode train \
+                --train /tmp/rk/currentData.csv \
+                --model /tmp/rk/results/simple_client \
+                --epochs 25 \
+                --embd 64 \
+                --hidden_char_dims 128 \
+                --domain_embd 32 \
+                --batch 256 \
+                --balanced_weights \
+                --model_output client \
+                --type rene
+
+python main.py --mode train \
+                --train /tmp/rk/currentData.csv \
+                --model /tmp/rk/results/simple_new_both \
+                --epochs 25 \
+                --embd 64 \
+                --hidden_char_dims 128 \
+                --domain_embd 32 \
+                --batch 256 \
+                --balanced_weights \
+                --model_output both \
+                --new_model \
+                --type rene
+
+python main.py --mode train \
+                --train /tmp/rk/currentData.csv \
+                --model /tmp/rk/results/simple_new_client \
+                --epochs 25 \
+                --embd 64 \
+                --hidden_char_dims 128 \
+                --domain_embd 32 \
+                --batch 256 \
+                --balanced_weights \
+                --model_output client \
+                --new_model \
+                --type rene

test.sh

@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_both --test /tmp/rk/futureData.csv --model_output both --type paul
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_client --test /tmp/rk/futureData.csv --model_output client --type paul
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_new_both --test /tmp/rk/futureData.csv --model_output both --type paul
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_new_client --test /tmp/rk/futureData.csv --model_output client --type paul
+
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_both --test /tmp/rk/futureData.csv --model_output both --type rene
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_client --test /tmp/rk/futureData.csv --model_output client --type rene
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_new_both --test /tmp/rk/futureData.csv --model_output both --type rene
+python main.py --mode test --batch 1024 --model /tmp/rk/results/simple_new_client --test /tmp/rk/futureData.csv --model_output client --type rene

visualize.py

@@ -2,7 +2,7 @@ import os
 
 import matplotlib.pyplot as plt
 import numpy as np
-from keras.utils import plot_model
+from keras.utils.vis_utils import plot_model
 from sklearn.decomposition import TruncatedSVD
 from sklearn.metrics import (
     auc, classification_report, confusion_matrix, fbeta_score, precision_recall_curve,