1: \begin{definition}[Relaxation sequence]

2: A relaxation sequence for $(\stdCone ,L,c)$ is a finite sequence 

3: of conic feasibility problems $\{ (\stdCone _{1},L,c), \ldots ,  (\stdCone _{\gamma},L,c)\}$ such that $K_1 = K$ and:

4: \begin{enumerate}

5: 	\item Every $K_i$ is an extended second order cone, see the beginning of Section \ref{sec:relaxation}.

6: 	\item For $i > 1$, there is $d^{i-1} \in K_{i-1}\cap L$ such that 

7: 	$(\stdCone _{i},L,c)$ is obtained as a result of applying Theorem \ref{theorem_reduction} to 

8: 	$K_{i-1}\cap L $ and $d^{i-1}$. In addition, we must have 

9: 	$K_{i-1} \subsetneq K_{i}$ (i.e., we do not admit trivial relaxations).

10: \end{enumerate}

11: The vectors in $\{d^{1}, \ldots , d^{\gamma -1} \}$ are called \emph{reducing directions}, due 

12: to the fact that they came from the application of facial reduction to the dual system $(\stdCone^*,L^\perp,0)$.

13: A relaxation sequence  is \emph{maximal} if it does not admit non-trivial relaxations.

14: The problem $(\stdCone _{\gamma},L,c)$ is called \emph{the last problem} of the sequence.

15: The length of the sequence is defined to be $\gamma$.

16: \end{definition}