aEIF psc models #513

Silmathoron · 2016-10-06T13:19:20Z

This PR adds implementations of the aEIF model with post-synaptic currents:

aeif_psc_alpha
aeif_psc_exp

Test is provided via an update of test_aeif_lsodar.py and by some of the usual suites such as issue-77.sli, which was modified.

…pikes)

…if_psc

golosio · 2016-10-15T15:59:32Z

@Silmathoron @heplesser if you agree I can review this PR

Silmathoron · 2016-10-15T16:16:50Z

Of course you can ;)
but I'm waiting for the VC on Monday to (most probably) remove the 2nd dynamics function and solve the conflicts

heplesser

@Silmathoron Except for a few small things, this looks fine to me.

heplesser · 2016-10-24T23:12:29Z

models/aeif_psc_alpha.cpp

@@ -94,19 +94,20 @@ nest::aeif_psc_alpha_dynamics( double,
  // good compiler will optimize the verbosity away ...

  // shorthand for state variables
-  const double& V = std::min( y[ S::V_M ], node.P_.V_peak_ ); // bound V
+  const double& V = std::min( y[ S::V_M ], node.P_.V_peak_ ); // bind V to the


I would suggest bound V by V_peak

I think this comment relates to an outdated version, current comment is
// bind V to the USER DEFINED V_peak_ value in Parameters.

@Silmathoron The English language is a challenging one: "bound" is both past tense of "bind" and a verb in its own right meaning "to confine" or "to limit". The original comment uses "bound" in the second meaning. "Bind" would be incorrect here, since it implied tying the membrane potential to V_peak, instead of just having V_peak as an upper limit. But this discussions shows the need for clearer language, so I would suggest: // enforce upper limit on V_m.

I would not mention "USER DEFINED V_peak_ value in Parameters", since that is clear from the code.

heplesser · 2016-10-24T23:13:08Z

models/aeif_psc_alpha.cpp

-  const double I_spike =
-    node.P_.Delta_T * std::exp( ( V - node.P_.V_th ) / node.P_.Delta_T );
+  const double I_spike = node.P_.Delta_T == 0. ? 0. : node.P_.g_L
+      * node.P_.Delta_T * std::exp( ( V - node.P_.V_th ) / node.P_.Delta_T );


For clarity, I would put the entire expression after : in parentheses, even though they are not strictly needed in C++.

heplesser · 2016-10-24T23:15:09Z

models/aeif_psc_exp.cpp

@@ -94,17 +94,18 @@ nest::aeif_psc_exp_dynamics( double, const double y[], double f[], void* pnode )
  // good compiler will optimize the verbosity away ...

  // shorthand for state variables
-  const double& V = std::min( y[ S::V_M ], node.P_.V_peak_ );
+  const double& V = std::min( y[ S::V_M ], node.P_.V_peak_ ); // bind V to the
+  // USER DEFINED V_peak_ value in Parameters.


See suggestion or alpha above.

heplesser · 2016-10-24T23:15:32Z

models/aeif_psc_exp.cpp

-  const double I_spike =
-    node.P_.Delta_T * std::exp( ( V - node.P_.V_th ) / node.P_.Delta_T );
+  const double I_spike = node.P_.Delta_T == 0. ? 0. : node.P_.g_L
+      * node.P_.Delta_T * std::exp( ( V - node.P_.V_th ) / node.P_.Delta_T );


Also here parentheses around all afer : would be helpful.

heplesser · 2016-10-25T08:05:21Z

@golosio I am happy with the code, how about you?

golosio · 2016-10-25T14:20:24Z

I have just very little suggestions/doubts

golosio · 2016-10-25T14:22:15Z

models/aeif_psc_alpha.cpp

+
+  updateValue< double >( d, names::gsl_error_tol, gsl_error_tol );
+
+  if ( V_reset_ >= V_peak_ )


It seems that also here the check that V_peak_ > V_th is missing

golosio · 2016-10-25T14:27:27Z

models/aeif_psc_alpha.cpp

+ * ---------------------------------------------------------------- */
+
+nest::aeif_psc_alpha::Parameters_::Parameters_()
+  : V_peak_( 0.0 )    // mV, should not be larger that V_th+10


I've seen this comment in all aeif models, but I do not understand it. Why shouldn't V_peak be larger that V_th+10? Even looking at the default values, the default value of V_th is -50.4, while V_peak is zero, so the difference is larger than 10. Am I missing something?

I never noticed that! It looks completely irrelevant to me...
@heplesser any thought on that?

I think the idea here is that the larger the V_th to V_peak is, the worse does the exploding-exponential problem become. But I think we handle this well elsewhere now, so the comment can be removed.

golosio · 2016-10-25T14:33:53Z

models/aeif_psc_exp.cpp

+
+  updateValue< double >( d, names::gsl_error_tol, gsl_error_tol );
+
+  if ( V_reset_ >= V_peak_ )


See above about the check that V_peak > V_th

Right, I forgot about that, thanks for reminding me! ;)

Besides the very little issues above, everything looks fine to me.

Silmathoron · 2016-10-25T16:22:43Z

Ok, I readded the missing test to all aeif_*_alpha models. I guess it went missing during some rearrangement in the dynamics modification.

…hould not be larger that V_th+10" comment on V_peak_

golosio · 2016-10-25T17:13:11Z

@Silmathoron @heplesser everythng is ok for me, I think this PR can be merged

heplesser · 2016-10-25T17:37:13Z

@Silmathoron Thanks for putting the test back in! But I think we should allow V_peak == V_th. Also, I would write the error message as "V_peak >= V_th required.". That is closer to what we use elsewhere and explicitly names the parameters involved.

…all aeif models except beta_multisynapse

Silmathoron · 2016-10-25T20:42:06Z

Ok, done! @golosio I'll let you do the changes on the beta_multisynapse model to avoid conflicts

heplesser · 2016-10-26T13:40:32Z

@Silmathoron Thanks for your efforts, all is well now! Together with @golosio's 👍, I will merge.

Silmathoron and others added 7 commits September 29, 2016 11:42

added psc models and included them in test_aeif_lsodar.py

f1bafed

corrected failing tests

fa45974

corrected clang formatting

da400a5

updated issue-77.sli (increase connection weight to 80. to initiate s…

4639569

…pikes)

Merge branch 'master' of git://github.com/nest/nest-simulator into ae…

69de6be

…if_psc

updated test and synaptic current names

30c0822

corrected formating

55e9b35

Silmathoron added 2 commits October 18, 2016 09:17

removed dynamics_DT0

c2f5342

resolved conflicts

9897e3f

heplesser requested changes Oct 24, 2016

View reviewed changes

Silmathoron added 2 commits October 25, 2016 08:56

minor changes to format

fa475a0

corrected the comment on V upper limit

5eae86a

heplesser approved these changes Oct 25, 2016

View reviewed changes

golosio reviewed Oct 25, 2016

View reviewed changes

readded test for V_peak > V_th to all aeif_*_alpha models; removed "s…

c917a53

…hould not be larger that V_th+10" comment on V_peak_

heplesser mentioned this pull request Oct 25, 2016

Correct input handling for aeif_cond_beta_multisynapse #510 and Update aeif_cond_beta_multisynapse to new aeif numerics approach #514 #516

Merged

changed peak conditions to V_peak_ < V_th and associated message for …

5fb1dee

…all aeif models except beta_multisynapse

heplesser approved these changes Oct 26, 2016

View reviewed changes

heplesser merged commit a418160 into nest:master Oct 26, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

aEIF psc models #513

aEIF psc models #513

Silmathoron commented Oct 6, 2016 •

edited

Loading

golosio commented Oct 15, 2016

Silmathoron commented Oct 15, 2016

heplesser left a comment

heplesser Oct 24, 2016

Silmathoron Oct 25, 2016

heplesser Oct 25, 2016

Silmathoron Oct 25, 2016

heplesser Oct 24, 2016

heplesser Oct 24, 2016

heplesser Oct 24, 2016

heplesser commented Oct 25, 2016

golosio commented Oct 25, 2016

golosio Oct 25, 2016 •

edited

Loading

golosio Oct 25, 2016

Silmathoron Oct 25, 2016

heplesser Oct 25, 2016

golosio Oct 25, 2016

Silmathoron Oct 25, 2016

golosio Oct 25, 2016

Silmathoron commented Oct 25, 2016

golosio commented Oct 25, 2016

heplesser commented Oct 25, 2016

Silmathoron commented Oct 25, 2016

heplesser commented Oct 26, 2016


		updateValue< double >( d, names::gsl_error_tol, gsl_error_tol );

		if ( V_reset_ >= V_peak_ )

aEIF psc models #513

aEIF psc models #513

Conversation

Silmathoron commented Oct 6, 2016 • edited Loading

golosio commented Oct 15, 2016

Silmathoron commented Oct 15, 2016

heplesser left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

heplesser commented Oct 25, 2016

golosio commented Oct 25, 2016

golosio Oct 25, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Silmathoron commented Oct 25, 2016

golosio commented Oct 25, 2016

heplesser commented Oct 25, 2016

Silmathoron commented Oct 25, 2016

heplesser commented Oct 26, 2016

Silmathoron commented Oct 6, 2016 •

edited

Loading

golosio Oct 25, 2016 •

edited

Loading