Add displacement argument to the Lyapunov exponent computation #36

SebastianM-C · 2017-10-04T21:38:33Z

No description provided.

Update to the latest version

Datseris

if you write state + displacement you are adding d0 twice

Datseris · 2017-10-04T21:39:53Z

src/lyapunovs.jl

@@ -322,7 +321,7 @@ function lyapunov_full(ds::ContinuousDynamicalSystem, T = 10000.0; Ttr = 0.0,
  st1 = ds.state
  integ1 = ODEIntegrator(ds, T; diff_eq_kwargs=diff_eq_kwargs)
  integ1.opts.advance_to_tstop=true
-  ds.state = st1 .+ d0
+  ds.state = st1 .+ displacement(st1, d0)


This doesn't work. You are adding d0 twice.
It is supposed to be state = displacement(state, d0)

Datseris · 2017-10-04T22:06:30Z

src/lyapunovs.jl

@@ -254,7 +254,7 @@ function lyapunov(ds::ContinuousDynamicalSystem, T = 10000.0; Ttr = 0.0,
  st1 = ds.state
  integ1 = ODEIntegrator(ds, T; diff_eq_kwargs=diff_eq_kwargs)
  integ1.opts.advance_to_tstop=true
-  ds.state = st1 .+ d0
+  ds.state = st1 .+ displacement(st1, d0)


This is also incorrect.

Datseris

If it is not added yet, please add this new keyword to the documentation string as well.

Datseris · 2017-10-04T22:12:51Z

src/lyapunovs.jl

@@ -254,7 +254,7 @@ function lyapunov(ds::ContinuousDynamicalSystem, T = 10000.0; Ttr = 0.0,
  st1 = ds.state
  integ1 = ODEIntegrator(ds, T; diff_eq_kwargs=diff_eq_kwargs)
  integ1.opts.advance_to_tstop=true
-  ds.state = st1 .+ displacement(st1, d0)
+  ds.state = displacement.(st1, d0)


I think the dot should be at the equals, i.e. .= displacement

Now, the function itself definitely doesn't need a dot. It doesn't operate element-wise on an array, but takes an array and produces a new one.

So ds.state .= displacement(st1, d0)

I mean, in many cases the user will specify a function that would not make sense to be broadcasted. This sacrifices some speed but gives flexibility.

Plus this part of the code is not the bottleneck so speed wouldn't matter here.

IMPORTANT. `ContinuousDynamicalSystem` must become parametric type with parameter `{D}`!!!!

the user defined distance may not give exactly `d0` distance

(changes necessary for Lyapunov)

Revert to 42a1f4 and re-add default_rescale

Remove the rest of the `{D}`s

(changes have to be applied to discrete systems as well)

Datseris

apply rescale to discrete system lyapunov as well

lyapunov still doesn't work... some problem with the rescale function I guess.

ChrisRackauckas · 2017-10-07T23:46:26Z

src/lyapunovs.jl

-                  T::Real = 10000.0,
-                  return_convergence::Val{B} = Val{false};
+                  T::Real,
+                  return_convergence::Val{B} = Val{false}();


Don't need the () if you do return_convergence::Type{Val{B}} which is better accepted.

(lyapunov still doesnt work!)

…ement

Datseris · 2017-10-08T00:20:19Z

You can know actually use lyapunov finally.

However, now it gives wrong results:

ds = Systems.lorenz()
lyapunov(ds, 10000.0)

gives 1.0 ... instead of 0.9 ...

I dont know whats wrong but it has to be fixed before merging..

i was supposed to be one loop outer!!!

Plus I changed all tabs to be 4 spaces instead of 2

plus added tests for discrete systems. The PR is DONE.

Datseris · 2017-10-09T21:48:49Z

BOOOM Test pass this NIGHTMARE can finally be merged after draining my almost empty soul.

@SebastianM-C I have changed completely the interface so you might want to study the source code before you use it.

SebastianM-C and others added 3 commits October 4, 2017 22:09

Merge pull request #1 from JuliaDynamics/master

3b5c375

Update to the latest version

Merge pull request #2 from JuliaDynamics/master

a6e34f7

Update to the latest version

Add displacement argument to lyapunov and lyapunov_full.

781507e

SebastianM-C changed the title ~~Displacement~~ Add displacement argument to the Lyapunov exponent computation Oct 4, 2017

SebastianM-C mentioned this pull request Oct 4, 2017

Add flexibility in the displacement used in lyapunov #33

Closed

Fix mistake

0db8a61

Datseris requested changes Oct 4, 2017

View reviewed changes

SebastianM-C and others added 10 commits October 6, 2017 00:30

Add the new keyword to the docstring

d31005c

change behavior of lyapunov + keyword arguments

42a1f4f

IMPORTANT. `ContinuousDynamicalSystem` must become parametric type with parameter `{D}`!!!!

ad0 in lyapunov

8f2c70a

the user defined distance may not give exactly `d0` distance

added parameter type D to ContinuousDS

f8db85a

(changes necessary for Lyapunov)

added parameter type D to DynamicalSystem

8158b72

added parametric type D to discrete systems

4927317

added default rescale

a8e81db

Revert parametric type D commits

458853d

Revert to 42a1f4 and re-add default_rescale

Merge remote-tracking branch 'origin/displacement' into displacement

3761d76

Fix mistake

7cb9b54

Remove the rest of the `{D}`s

Datseris previously approved these changes Oct 7, 2017

View reviewed changes

SebastianM-C added 2 commits October 8, 2017 02:06

Remove {D}s

0f5805e

Fix more mistakes

a75698f

Datseris reviewed Oct 7, 2017

View reviewed changes

fix systems

139b347

lyapunov still doesn't work... some problem with the rescale function I guess.

ChrisRackauckas reviewed Oct 7, 2017

View reviewed changes

Datseris added 5 commits October 8, 2017 01:47

Val{B} to Type{Val{B}}

98247cb

(lyapunov still doesnt work!)

rescale(length) -> rescale(ds)

7780d18

fixin lyapunov...

3bbc3c1

fixin lyapunov...

7d832aa

Merge remote-tracking branch 'SebastianM-C/displacement' into displac…

66f1346

…ement

Datseris added 7 commits October 9, 2017 21:25

change default_rescale argument

a50a64d

fix humongous error with i

8380144

i was supposed to be one loop outer!!!

displacement working for discrete

9413761

continuous displacement works

9ec33b1

Plus I changed all tabs to be 4 spaces instead of 2

updated lyapunov continuous tests

a69b825

added option to choose convergence for discrete

c7423ab

plus added tests for discrete systems. The PR is DONE.

fi retarded test typo

53e5449

Datseris merged commit 46f6129 into JuliaDynamics:master Oct 9, 2017

Datseris mentioned this pull request Oct 9, 2017

Add possibility to give lyapunov convergence timeseries instead of only final value #18

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add displacement argument to the Lyapunov exponent computation #36

Add displacement argument to the Lyapunov exponent computation #36

SebastianM-C commented Oct 4, 2017

Datseris left a comment

Datseris Oct 4, 2017

Datseris Oct 4, 2017

Datseris left a comment

Datseris Oct 4, 2017

Datseris Oct 4, 2017

Datseris left a comment

ChrisRackauckas Oct 7, 2017

Datseris commented Oct 8, 2017

Datseris commented Oct 9, 2017

Add displacement argument to the Lyapunov exponent computation #36

Add displacement argument to the Lyapunov exponent computation #36

Conversation

SebastianM-C commented Oct 4, 2017

Datseris left a comment

Choose a reason for hiding this comment

Datseris Oct 4, 2017

Choose a reason for hiding this comment

Datseris Oct 4, 2017

Choose a reason for hiding this comment

Datseris left a comment

Choose a reason for hiding this comment

Datseris Oct 4, 2017

Choose a reason for hiding this comment

Datseris Oct 4, 2017

Choose a reason for hiding this comment

Datseris left a comment

Choose a reason for hiding this comment

ChrisRackauckas Oct 7, 2017

Choose a reason for hiding this comment

Datseris commented Oct 8, 2017

Datseris commented Oct 9, 2017