The most suitable deep learning model for sequential data classification, with a focus on achieving high accuracy, low latency, and small size, is the Long Short-Term Memory (LSTM) or GRU (Gated Recurrent Units) networks. These networks are specifically designed to process sequential data, such as text and time signals, as they allow information to be retained over long periods of time. LSTM and GRU are able to adapt to changes in data and provide fast response, which helps in improving latency. Moreover, GRU is less complex in terms of the number of units compared to LSTM, resulting in a smaller and faster model to train. Ultimately, both LSTM and GRU are excellent choices for classifying sequential data effectively and efficiently.

For sequential data classification, where the goal is high accuracy, low latency, and small model size, Recurrent Neural Networks (RNNs), specifically Long Short-Term Memory (LSTM) or Gated Recurrent Units (GRU), are commonly used.

Recurrent Neural Networks (RNNs) work well with sequential data. However, you have to check for Vanishing and Exploding Gradients.

Article Efficient Deep Learning: A Survey on Making Deep Learning Mo...