Code

[MRG] Efficient synapse creation#600

Code

Merged

mstimberg merged 84 commits intomasterbrian-team/brian2:masterfrom efficient_synapse_creationbrian-team/brian2:efficient_synapse_creationCopy head branch name to clipboardApr 9, 2016

+2,705-1,043

Commits

Commits on Nov 24, 2015

• Some ideas for analysing synaptic conditions

  thesamovar

  committed
  2b0839b

Commits on Nov 25, 2015

• Improvements

  thesamovar

  committed
  6cb897b

Commits on Dec 18, 2015

• Merge remote-tracking branch 'origin/master' into efficient_synapse_creation

  thesamovar

  committed
  fe6c53b

Commits on Dec 19, 2015

• Probably breaking change implementing efficient binomial synapse creation algorithm

  thesamovar

  committed
  4acc629

Commits on Dec 24, 2015

• Parsing generator syntax

  Show description for f3e5b65
  thesamovar

  committed
  f3e5b65

• Improvements to generator syntax checking function

  Show description for e3d5ab8
  thesamovar

  committed
  e3d5ab8

Commits on Jan 31, 2016

• Merge branch 'master' into efficient_synapse_creation

  thesamovar

  committed
  8cd5feb

• Won't use conditions

  thesamovar

  committed
  732c8f9

• Working synapse generator syntax parser (and tests)

  thesamovar

  committed
  bda0e77

• Change num keyword to size

  thesamovar

  committed
  110655c

• Working version with improved numpy synapse construction speed

  thesamovar

  committed
  bef406d

Commits on Mar 10, 2016

• Merge remote-tracking branch 'remotes/origin/master' into efficient_synapse_creation

  mstimberg

  committed
  7b40052

Commits on Mar 17, 2016

• Do not use random sampling if p = 1

  mstimberg

  committed
  8a36075

• Move _sample_without_replacement out of the numpy template so that it can also be used by weave/cython

  Show description for f64ce38
  mstimberg

  committed
  f64ce38

• Remove the previous way of generating random connections from the numpy template (was never used in the current logic)

  mstimberg

  committed
  5b57dd9

• Also move the binomial function out of the template

  mstimberg

  committed
  53d7a9c

• Implement the more efficient probabilistic synapse creation for weave

  mstimberg

  committed
  e5e5a70

• Implement the more efficient probabilistic synapse creation for Cython

  mstimberg

  committed
  7a67ae1

• Fix synapse creation for subgroups

  mstimberg

  committed
  f200f98

Commits on Mar 23, 2016

• Implement a new logic for choosing a sampling algorithm (stochastic synapse creation)

  Show description for 8929cd4
  mstimberg

  committed
  8929cd4

• Fix: do not use the "rand()<p" approach when already using the "sampling" approach

  mstimberg

  committed
  f3ccbbf

Commits on Mar 24, 2016

• Quick fix to make parse_synapse_generator compatible with Python 3.5

  mstimberg

  committed
  7fd7a88

• Fix certain combinations of conditions and probabilities in the probabilistic synapse generation with numpy

  Show description for 36d6ef5
  mstimberg

  committed
  36d6ef5

Commits on Apr 1, 2016

• Merge remote-tracking branch 'origin/master' into efficient_synapse_creation

  thesamovar

  committed
  8f64dd0

• Typo correction in docs

  Show description for 55e2b8c
  thesamovar

  committed
  55e2b8c

Commits on Apr 2, 2016

• First version of generator syntax working for some examples (e.g. CUBA works, COBA doesn't)

  thesamovar

  committed
  09381c4

• Improvements to implementation of generator syntax (still unlikely to be working in all cases I think)

  thesamovar

  committed
  4925b03

• Implemented n for generator syntax, almost all tests passing except for string syntax for n, p

  thesamovar

  committed
  9bcaca1

• Attempt to deal with vectorisation idx (but not working still)

  thesamovar

  committed
  f36719c

Commits on Apr 3, 2016

• Allow for string p and n (but not p dependent on postsynaptic index)

  thesamovar

  committed
  a133593

• All tests pass for numpy

  thesamovar

  committed
  c9845fa

• All tests pass for weave

  thesamovar

  committed
  61c8bb2

• Jump optimisation for weave

  thesamovar

  committed
  d6c3b79

• Better optimisation

  thesamovar

  committed
  9bd7f5f

• Cython tests now passing (although needed a hack to make it work). All non-standalone tests should be passing now I guess

  thesamovar

  committed
  7994007

• Standalone implemented, but CUBA and COBA examples crash

  thesamovar

  committed
  bb3e302

• Fixed a couple of standalone tests and added check for fixed sample size

  thesamovar

  committed
  4dd3a38

Commits on Apr 4, 2016

• Fixed _vectorisation_idx issue

  thesamovar

  committed
  08fd8cf

Commits on Apr 5, 2016

• Fix incorrect use of {..} in Synapses docstrings

  mstimberg

  committed
  2727f36

• Add more tests for the new generator syntax in Synapses.connect

  mstimberg

  committed
  53b8126

• Make the test suite fail by default if a test raises a NotImplementedError

  mstimberg

  committed
  76b6c56

• Make several of the synapse connection tests "standalone-comaptible"

  mstimberg

  committed
  bb97b5e

• Print the project directory for standalone run failures

  mstimberg

  committed
  a637ab6

• Prevent segmentation faults in standalone for empty Synapses objects

  mstimberg

  committed
  cdf255f

• Raise error message for generator expression as condition, and fix weird weave bug

  thesamovar

  committed
  2d78d90

• Minor fix to test_synapses.py

  thesamovar

  committed
  afeb992

• Turn off restricted pointers for scalar values for safety

  thesamovar

  committed
  b2482a6

• Optimisation/fix for iteration variable dependence in condition

  thesamovar

  committed
  cc28fb9

• Merge remote-tracking branch 'origin/master' into efficient_synapse_creation

  thesamovar

  committed
  a8ff0bf

• Support for int arguments to pow for MSVC

  thesamovar

  committed
  c1b1a9c

• Fixed standalone crash problem

  thesamovar

  committed
  db70a3a

Commits on Apr 6, 2016

• Documentation and examples updated

  thesamovar

  committed
  bfab214

• Add missing reinit_devices teardown to tests marked as "standalone-compatible"

  mstimberg

  committed
  863e199

• No need to install miniconda on appveyor, it is already pre-installed

  Show description for 487ce48
  mstimberg

  committed
  487ce48

• Fix(?) the path to anaconda on appveyor

  mstimberg

  committed
  84f5b52

• ... and do not use an invalid appveyor.yml in the first place...

  mstimberg

  committed
  ecc99da

• Try setx instead of set to set the path and add some more debug output

  mstimberg

  committed
  7e362c3

• ...trying to correct the setx use

  mstimberg

  committed
  c329112

• ... trying to fix setx once more...

  mstimberg

  committed
  0e1838b

• ...sigh... actually use the Miniconda (and not the Python) paths; revert most of the previous changes to appveyor.yml

  mstimberg

  committed
  8ccef39

• Various documentation updates and fixed bug with rand()<p introduced by previous optimisation

  thesamovar

  committed
  4bd9869

• Another small change to docs

  thesamovar

  committed
  592d800

• Add block names to LIO constants (fixes #662)

  thesamovar

  committed
  e06a390

• Fixed tests to not use connect=

  thesamovar

  committed
  fc5b965

Commits on Apr 7, 2016

• Add more tests for synapse creation with subgroups (and make them "standalone-compatible")

  mstimberg

  committed
  19aede6

• Bugfix and numpy optimisation

  thesamovar

  committed
  20d2a92

• Check indices are valid for generator syntax

  thesamovar

  committed
  5781352

• Fix for standalone

  thesamovar

  committed
  6fabf1d

• Small fix to skip one part of a test for standalone

  thesamovar

  committed
  198e9d9

• Merge remote-tracking branch 'origin/master' into efficient_synapse_creation

  thesamovar

  committed
  bea3fe6

Commits on Apr 8, 2016

• Use str(e) instead of e.message (the latter does not work in Python 3)

  mstimberg

  committed
  674f2c3

• Avoid that a synapse connection test generates invalid j values

  mstimberg

  committed
  d8dfb78

• Fix for presynaptic condition not evaluated outside range in standalone

  thesamovar

  committed
  8321644

• Added new skip_if_invalid syntax

  thesamovar

  committed
  2e32809

• Slightly more comprehensive testing for multiple synapses

  mstimberg

  committed
  e9ba301

• Some cleaning up, stylistic changes to the code

  mstimberg

  committed
  a49f9ef

• Fix synapse generation in numpy for n=0

  mstimberg

  committed
  9d53aeb

• Add an error message for Synapses(..., connect=...)

  mstimberg

  committed
  2e34b7d

• More tests for multiple synapses, in particular for subgroups

  mstimberg

  committed
  4c1353a

• Fix skip_if_invalid for numpy

  mstimberg

  committed
  6fe9d3e

• Remove a TODO

  Show description for 932a70b
  mstimberg

  committed
  932a70b

• More specific error message for incorrect first argument of Synapses.connect

  Show description for b3d95aa
  mstimberg

  committed
  b3d95aa

• Re-enable objects that have indices (e.g. morphologies) as arguments for i and j in Synapses.connect

  mstimberg

  committed
  53e724d

• Added test of statistical distribution of Synapse connectivity with p=...

  Show description for 4a1cdc9
  thesamovar

  committed
  4a1cdc9