1: \begin{abstract}

2: As the number of sensors becomes massive in Internet of Things (IoT) networks, the amount of data is humongous.

3: %

4: To process data in real-time while protecting user privacy, federated learning (FL) has been regarded as an enabling technique to push edge intelligence into IoT networks with massive devices.

5: %

6: However, FL latency increases dramatically due to the increase of the number of parameters in deep neural network and the limited computation and communication capabilities of IoT devices.

7: %

8: To address this issue, we propose a semi-federated learning (SemiFL) paradigm in which network pruning and over-the-air computation are efficiently applied.

9: %

10: To be specific, each small base station collects the raw data from its served sensors and trains its local pruned model. 

11: %

12: After that, the global aggregation of local gradients is achieved through over-the-air computation.

13: %

14: We first analyze the performance of the proposed SemiFL by deriving its convergence upper bound.

15: %

16: To reduce latency, a convergence-constrained SemiFL latency minimization problem is formulated.

17: %

18: By decoupling the original problem into several sub-problems, iterative algorithms are designed to solve them efficiently.

19: %

20: Finally, numerical simulations are conducted to verify the effectiveness of our proposed scheme in reducing latency and guaranteeing the identification accuracy.

21: \end{abstract}