It has a good treatment of recurrent neural networks, which is nice since the theory behind them starts being relevant again. For comparison, both Bishop's new

It has a good treatment of recurrent neural networks, which is nice since the theory behind them starts being relevant again. For comparison, both Bishop's new deep learning book and Simon Price's Understanding deep learning get straight to transformers and only mention RNN/LSTM in passing - they were written just as state space models and hybrid RNN/transformer models started showing good results.