I have a vector as a list, x = [a1,a2,a3,…aN]. I pass it through a fully connected layer defined as this
<code>dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
</code>
<code>dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
</code>
dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
This fully connected layer is not trained. I am curious to whether the output
<code>output = dense(tf.constant([x]))
</code>
<code>output = dense(tf.constant([x]))
</code>
output = dense(tf.constant([x]))
would represent the initial vector x in any sort of way despite the untrained dense layer?
I am aware that every time that run
<code>x = [a1,a2,a3,...aN]
dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
output = dense(tf.constant([x]))
</code>
<code>x = [a1,a2,a3,...aN]
dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
output = dense(tf.constant([x]))
</code>
x = [a1,a2,a3,...aN]
dense = tf.keras.layers.Dense(units=8, kernel_initializer='glorot_uniform')
output = dense(tf.constant([x]))
the output would be different. However, if I establish the fully connected layer, then pass many vector through the layer, would my outputs be a representation of the vectors I inputed, or random nonsense?