- I am building binary classification model using neural network.
- All features are categorical variables, hence I am using nn.embedding layer. Features are [‘action’, ‘subtype_action’, ‘user’] . I have written model at the end the query.
- I need to update this model every day and each day new users will come. Let us say maximum number of users are 50K for lifetime of a model and each day 1K new users join.
- In that case can I add embedding layer of (Max_num_users, embedding_size_for_users) from day one. Or else is there any better way to increase embedding layer 1K each day.
Below is my model
BinaryClassification(
(emb_layers): ModuleList(
(0): Embedding(4, 2)
(1): Embedding(18, 2)
(2): Embedding(1000, 8)
)
(linear_layers): ModuleList(
(0): Linear(in_features=14, out_features=8, bias=True)
)
(output_layer): Linear(in_features=8, out_features=1, bias=False