Forget gate Deep Learning
noun phrase
Definition: In an LSTM cell, the gate that regulates how much information from the previous cell state is retained or discarded at the current time step. It typically uses a sigmoid activation to produce values between 0 and 1, thereby controlling the extent to which earlier memory is preserved [Chambers et al. 2024].
Example in context: “The first gate, a forget gate, employs a sigmoid layer to determine which cell state information to eliminate.” [Sama, Hashmi 2024]
Related terms: LSTM gate, input gate, output gate, cell state