1: \begin{abstract}

2: Many real-world processes evolve in cascades over complex networks, whose topologies 

3: are often unobservable and change over time. However, 

4: the so-termed adoption times when blogs mention popular news items, 

5: individuals in a community catch an infectious disease, or 

6: consumers adopt a trendy electronics product are typically known,

7: and are implicitly dependent on the underlying network. 

8: To infer the network topology, a \textit{dynamic} 

9: structural equation model is adopted to capture the 

10: relationship between observed adoption times and the unknown edge weights. 

11: Assuming a slowly time-varying topology and leveraging the 

12: sparse connectivity inherent to social networks, 

13: edge weights are estimated by minimizing a sparsity-regularized 

14: exponentially-weighted least-squares criterion. 

15: To this end, solvers with complementary strengths 

16: are developed by leveraging (pseudo) real-time 

17: sparsity-promoting proximal gradient iterations, 

18:  the improved convergence rate of accelerated variants,

19: or reduced computational complexity of stochastic gradient descent.

20: Numerical tests with both synthetic and real data demonstrate the effectiveness of the 

21: novel algorithms in unveiling sparse dynamically-evolving 

22: topologies, while accounting for external influences in the adoption times. 

23: Key events in the recent succession of political leadership in North Korea, explain 

24: connectivity changes observed in the associated network inferred from global cascades

25: of online media.

26: \end{abstract}

27: