@@ -39,6 +39,7 @@ The hierarchy of dense matrices provided are
39
39
- Circular symplectic ensemble (GSE, ` beta=4 ` )
40
40
- Laguerre matrices = white Wishart matrices
41
41
- Jacobi matrices = MANOVA matrices
42
+ - Unitary invariant ensembles
42
43
43
44
Unless otherwise specified, ` beta=1,2,4 ` are supported. For the symplectic matrices ` beta=4 ` ,
44
45
the 2x2 outer block-diagonal complex representation ` USp(2N) ` is used.
@@ -82,6 +83,11 @@ Hermite, Laguerre(m) and Jacobi(m1, m2) ensembles.
82
83
is applied to the raw QR decomposition. By default, ` correction=1 ` (Edelman's correction) is
83
84
used. Other valid values are ` 0 ` (no correction) and ` 2 ` (Mezzadri's correction).
84
85
- ` NeedsPiecewiseCorrection() ` implements a simple test to see if a correction is necessary.
86
+
87
+ - ` InvariantEnsemble(str,n) `
88
+ Generates a unitary invariant ensemble, where str determines the
89
+ potential of the ensemble, see below.
90
+ Only available if ApproxFun package is installed.
85
91
86
92
The parameters ` m ` , ` m1 ` , ` m2 ` refer to the number to independent "data" degrees of freedom.
87
93
For the dense samples these must be ` Integer ` s but can be ` Real ` s for the rest.
@@ -146,6 +152,23 @@ In the following, `dt` is the time interval being discretized over and `t_end` i
146
152
- ` StochasticAiryProcess(dt, t_end, beta) ` generates the largest eigenvalue corresponding to the
147
153
stochastic Airy process with real positive ` beta ` . This is known to be distributed in the ` t_end -> Inf `
148
154
limit to the ` beta ` -Tracy-Widom law.
155
+
156
+ # Invariant ensembles
157
+
158
+ ` InvariantEnsemble(str,n) ` supports n x n unitary invariant ensemble
159
+ with distribution
160
+
161
+ ` exp(- Tr Q(M)) dM `
162
+
163
+ ` str ` specifies an ensemble with precomputed recurrence coefficients.
164
+ The currently include ensembles are
165
+
166
+ - Quartic: Q(M) = n M^4
167
+ - Eight: Q(M) = n M^8
168
+ - HODecay: Q(M) = n (M^4/20 - 4/15M^3 +M^2/5 + 8/5M)
169
+ - CoshUnscaled: Q(M) = cosh(M)
170
+ - QuarticUnscaled: Q(M) = M^4
171
+ - EightUnscaled: Q(M) = M^8
149
172
150
173
# References
151
174
- James Albrecht, Cy Chan, and Alan Edelman,
0 commit comments