前几天去厦门开会(DDAP10),全英文演讲加之大家口音都略重,说实话听演讲主要靠看ppt,摘出一篇听懂的写篇博客纪念一下吧。
11.2 Session-A 13:30-18:00 WICC G201
Time | Speaker | No. | Title |
---|---|---|---|
14:30-15:00 | Wei Lin | ST-07 | Dynamical time series analytics: From networks construction to dynamics prediction |
主要讲了他的两个工作,一个是重构的工作,一个是预测的工作,分别发表在PRE和PNAS上。
第一篇工作
Detection of time delays and directional interactions based on time series from complex dynamical systems
ABSTRACT
Data-based and model-free accurate identification of intrinsic(固有) time delays and directional interactions.
METHOD
Given a time series x ( t ) x(t) x(t), one forms a manifold(流形) M X ∈ R n M_X\in R^n MX∈Rn based on delay coordinate embedding: X ( t ) = [ x ( t ) , x ( t − δ t ) , . . . , x ( t − ( n − 1 ) δ t ) ] X(t) = [x(t),x(t − \delta t), . . . ,x(t − (n − 1)\delta t)] X(t)=[x(t),x(t−δt),...,x(t−(n−1)δt)], where n n n is the embedding dimension and δ t \delta t δt is a proper time lag.
CME method:
Say we are given time series x ( t ) x(t) x(t) and y ( t ) y(t) y(t) as well as a set of possible time delays: Γ = { τ 1 , τ 2 , … , τ m } \Gamma = \{\tau_1,\tau_2, … ,\tau_m\} Γ={ τ1,τ2,…,τm}. For each candidate time delay τ i \tau_i τi, we let z ( t ) = x ( t − τ i ) z(t) = x(t − \tau_i) z(t)=x(t−τi) and form the manifolds M Y M_Y MY and M Z M_Z MZ with n y n_y ny and n z n_z nz being the respective embedding dimensions. For each point Y ( t ^ ) ∈ M Y Y(\hat{t}) \in M_Y Y(t^)∈MY , we find K K K nearest neighbors Y ( t j ) ( j = 1 , 2 , … , K ) Y(t_j)(j = 1,2, …,K) Y(tj)(j=1,2,…,K), which are mapped to the mutual neighbors Z ( t j ) ∈ M Z ( j = 1 , 2 , … , K ) Z(t_j) \in M_Z(j = 1,2, …,K) Z(tj)∈MZ(j=1,2,…,K) by the cross map. We then estimate Z ( t ) Z(t) Z(t) by averaging these mutual neighbors through Z ^ ( t ^ ) ∣ M Y = ( 1 / K ) ∑ j = 1 K Z ( t j ) \hat{Z}(\hat{t})|M_Y=(1/K)\sum^K_{j=1}Z(t_j) Z^(t^)∣MY=(1/K)∑j=1KZ(tj). Finally, we define the CME score as
s ( τ ) = ( n Z ) − 1 t r a c e ( Σ Z ^ − 1 c o v ( Z ^ ,