![\"Dimensionality](\"../../support/images/DRDL.png\")
\n",
"\n",
"\n",
"\n",
"The 3-dimensional representations of $h_{t}$ ($\\hat{h_{t}}$) for both stimulus class are plotted along the three axes of the coordinate system below. The plot represents the temporal trajectories of the two classes. At the first timepoint the two classes are closest to each other. Distance between them increases with every timepoint. Next plot shows the Euclidean distance between the two classes as a function of time."
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {
"tags": [
"remove-cell"
]
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"WARNING:tensorflow:Layer gru will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
"WARNING:tensorflow:Layer gru_1 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
"WARNING:tensorflow:Layer gru_2 will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
"Model: \"sequential\"\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"gru (GRU) (None, None, 16) 16032 \n",
"_________________________________________________________________\n",
"gru_1 (GRU) (None, None, 16) 1632 \n",
"_________________________________________________________________\n",
"gru_2 (GRU) (None, None, 16) 1632 \n",
"_________________________________________________________________\n",
"time_distributed (TimeDistri (None, None, 3) 51 \n",
"=================================================================\n",
"Total params: 19,347\n",
"Trainable params: 0\n",
"Non-trainable params: 19,347\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"# Make a copy the model\n",
"# (tips: https://stackoverflow.com/questions/57316557/tf-keras-layers-pop-doesnt-work-but-tf-keras-layers-pop-does)\n",
"#tf.random.set_seed(50)\n",
"tf.random.set_seed(47)\n",
"BestModel = tf.keras.models.clone_model(model)\n",
"BestModel.set_weights(model.get_weights())\n",
"\n",
"# Replace the the last Time-Distributed layer with Fully connected layer with only two units\n",
"BestModel = tf.keras.Sequential(BestModel.layers[:-1])\n",
"\n",
"# add a new time-distributes dense layer with 2 units and 'linear' activation function\n",
"BestModel.add(tf.keras.layers.TimeDistributed(tf.keras.layers.Dense(3,activation='linear')))\n",
"\n",
"# Set trainable = False for the layers from BestModel\n",
"for layers in BestModel.layers:\n",
" layers.trainable = False\n",
"\n",
"BestModel.summary()"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {
"tags": [
"remove-cell"
]
},
"outputs": [],
"source": [
"trajectory = {}\n",
"for subj_idx in dataset.test_idx:\n",
" subj = dataset.sid()[subj_idx]\n",
" X_test, y_test = query_dataset(dataset_df,[subj_idx])\n",
" y_pred = BestModel.predict(X_test)\n",
" appr = y_pred[y_test==1.,:,:].mean(axis=0)\n",
" retr = y_pred[y_test==0.,:,:].mean(axis=0)\n",
" trajectory[subj] = {'Approach':{'x1':list(appr[:,0]),'x2':list(appr[:,1]),'x3':list(appr[:,2])},\n",
" 'Retreat':{'x1':list(retr[:,0]),'x2':list(retr[:,1]),'x3':list(retr[:,2])}}"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {
"tags": [
"remove-cell"
]
},
"outputs": [],
"source": [
"trajectory_df = pd.DataFrame.from_dict({(i,j,k): trajectory[i][j][k] \n",
" for i in trajectory.keys() \n",
" for j in trajectory[i].keys() \n",
" for k in trajectory[i][j].keys()}).T\n",
"trajectory_df = trajectory_df.stack().unstack(2).stack()\n",
"trajectory_df = trajectory_df.to_frame(name='trajectory')\n",
"trajectory_df.reset_index(inplace=True)\n",
"trajectory_df.rename(columns={'level_0':'Subj','level_1':'Direction','level_2':'TP','level_3':'axis'},inplace=True)\n",
"trajectory_df = trajectory_df.groupby(['Direction','TP','axis'])['trajectory'].mean().unstack(-1).reset_index()"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {
"tags": [
"remove-cell"
]
},
"outputs": [],
"source": [
"approach_x1 = trajectory_df[trajectory_df['Direction']=='Approach']['x1']\n",
"approach_x2 = trajectory_df[trajectory_df['Direction']=='Approach']['x2']\n",
"approach_x3 = trajectory_df[trajectory_df['Direction']=='Approach']['x3']\n",
"\n",
"retreat_x1 = trajectory_df[trajectory_df['Direction']=='Retreat']['x1']\n",
"retreat_x2 = trajectory_df[trajectory_df['Direction']=='Retreat']['x2']\n",
"retreat_x3 = trajectory_df[trajectory_df['Direction']=='Retreat']['x3']"
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {
"tags": [
"remove-input"
]
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
"\n",
"\n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
"
\n",
"\n",
""
],
"text/plain": [
"