Software:Release Notes 2.0

From NEST

Version 2.0 RC4 (2011-03-24)

• Fixed installation of the topology testsuite.
• spike_generator now checks all spike times for grid-compatibility when setting them. This avoids crashes during simulations in multi-threaded scenarios.
• The Hill-Tononi model (ht_neuron) now is absolutely refractory while the spike current g_spike is active. This is the correct behavior after careful re-reading of the paper and inspection of the original source code.
• PyNEST now works again with Python 2.4
• MyModule now builds and installs again

Version 2.0 RC3 (2011-02-25)

Improvements and bug fixes

• Multiple multimeterss can now be connected to a neuron (thanks to Pierre Yger for pointing out this problem).
• multimeter now has an accumulator mode, summing values across all recorded nodes for each time step.
• New SLI functions DoubleQ, IntegerQ to check numbers for being of type double, integer, respectively.
• New SLI function FiniteQ to check if an object is a finite number.
• Set the precise_times flag of spike_detector to true and precision to 15 if precise spiking neuron model are in the network and the user did not set this properties explicitly.
• PyNEST voltage_trace now works together with multimeter.
• PyNEST can now be built without NumPy. Please note, however, that this leads to a certain performance penalty.

Documentation

• New and extended documentation on documentation on parallel and distributed computing.
• Small corrections and new examples in the topology user manual.

Bug fixes

• Re-added MUSIC support, which got lost in an earlier release. See Using NEST with MUSIC.
• SLI functions shrink and reserve now work as expected.
• PyNEST testsuite now correctly shows the number of errors and failures.
• PyNEST raster_plot / voltage_trace now correctly handles parameters and grayscale plotting again.

Known issues

• Two test of the built-in testsuite currently fail. The reason for this is known and this will be fixed soon in RC4.

Version 2.0 RC2 (2010-11-26)

• Topology: Change semantics of anchor supplied with kernel functions: anchor now gives displacement relative to driver neuron, as for free masks.
• Topology plotting functions improved now respect mask and kernel anchors and have a more flexible interface
• New Topology user manual (PDF: Image:Topology UserManual 2.0-RC2.pdf)
• Modified topology_func() to be Python 2.5 compatible
• Corrected memory leak in spike_generator for sending weighted spikes.

Version 2.0 RC1 (2010-11-11)

New functionality

• New models stdp_dopamine_synapse and volume_transmitter to add support for dopamine-modulated spike-timing dependent plasticity, as described in Potjans W, Morrison A and Diesmann M (2010) Enabling functional neural circuit simulations with distributed computing of neuromodulated plasticity', Front. Comput. Neurosci. 4:141.
• New model: iaf_psc_exp_ps, a leaky integrate-and-fire neuron with exponential postsynaptic currents (canoncial implementation). It uses the bisectioning method for the approximation of the threshold crossing. See Hanuschkin A, Kunkel S, Helias M, Morrison A and Diesmann M (2010) A general and efficient method for incorporating precise spike times in globally time-driven simulations, Front. Neuroinform 4:113.
• ConnPlotter is now part of the official NEST release. For more information on ConnPlotter, see Nordlie E, Plesser HE (2009) Visualizing neuronal network connectivity with connectivity pattern tables, Front Neuroinform 3:39.
• NEST now supports runtime interoperability with other applications via the MUSIC interface. For details on the design, the usage and the performance of the interface, see Djurfeldt M, Hjorth J, Eppler JM, Dudani N, Helias M, Potjans TC, Bhalla US, Diesmann M, Kotaleski JH and Ekeberg Ö (2010) Run-Time Interoperability Between Neuronal Network Simulators Based on the MUSIC Framework, Neuroinformatics 8(1):43-60. Please note that since the publication of the article, we changed the name of music_{in,out}_proxy to music_event_{in,out}_proxy and added support for receiving continous data and string messages. For examples, please see Using NEST with MUSIC.
• New model: sli_neuron for rapid prototyping of neuron and device models in SLI.
• New SLI function Rotate for arrays

  SLI ] [-2 -1 0 1 2] 2 Rotate ==
  [0 1 2 -2 -1]
  SLI ] [-2 -1 0 1 2] -2 Rotate ==
  [1 2 -2 -1 0]

• New SLI functions for strings: ToUppercase, ToLowercase

  SLI ] (MiXeD CaSe) ToUppercase dup ==
  (MIXED CASE)
  SLI [1] ToLowercase ==
  (mixed case)

• SLI function Sort now also supports arrays of strings

  SLI ] [(a) (A) (b) (C)] Sort == 
  [(A) (C) (a) (b)]

• New SLI functions for getting local nodes: GetLocalLeaves, GetLocalNodes, GetLocalChildren
• New SLI function ScanThread to execute a function to corresponding elements of n arrays.
• New SLI function StringQ for string type testing.
• The ht_generator has been renamed smp_generator (sinusoidally modulate Poisson generator).

Improvements

• Devices now collect data across all threads. This eliminates the need for extended addresses and loops over the different threads to collect the data internally. The following listing illustrates the old way of retrieving the membrane potential from a voltmeter in a simulation with 4 threads:

  [0 1 2 3] { vm GetAddress exch append GetStatus /events get /V_m get} Map Flatten

  The new way to retrieve the data looks like this:

  vm GetStatus /events get /V_m get

  or using the hierarchical get operator in short:

  vm [/events /V_m] get 

  The only change for user code is that the property filename of all recording devices has been renamed to filenames and contains one filename per thread now.
• All neuron models now support multimeter. Support for all other analog recorders has been removed. voltmeter has been added as multimeter preconfigured to record voltages.
  • User documentation: Analog recording with multimeter
  • Developer documentation: Converting models to support multimeter recording
• NEST now invokes MPI_Abort() in case of errors in the script. This means that errors terminate all connected processes, which minimizes the danger of dead-locks on the cluster.
• Improved PyNEST's RandomConvergentConnect and RandomDivergentConnect by moving loops to SLI and by adding options to allow/disallow autapses/multapses from within PyNEST. See PyNEST#Connections for help on the usage.
• First tests for MPI. If you are using distributed computing and want to run the testsuite, you have to add the function /mpirun to the NEST config file ~/.nestrc. Please see $PREFIX/share/doc/nest/examples/nestrc.sli for an example implementation.
• Cleanup of the benchmarks in Brette_et_al to conform to the new API and produce consistent results.
• PyNEST submodule voltage_trace now supports voltmeters that recorded from multiple neurons.
• Added raw count histogram to correlation_detector.
• GSL error tolerance now configurable for aeif models, and models throw exceptions on implausible state values
• Improved performance for communicating off-grid spikes in distributed setups

Improved Topology Module

The Topology Module has been improved, including some changes that change the behavior.

• The Topology PyNEST user interface is now much closer to the general PyNEST interface.
• ConnectLayer is now called ConnectLayers
• All nest.topology functions now require lists of GIDs as input, not "naked" GIDs
• There are a number of new functions in nest.topology, I tried to write good doc strings for them
• For grid based layers (ie those with /rows and /columns), we have changed the definition of extent: Previously, nodes were placed on the edges of the extent, so if you had an extend of 2 (in x-direction) and 3 nodes, these had x-coordinates -1, 0, 1. The grid constant was extent/(num_nodes - 1). Now, we define the grid constant as extent/num\_nodes, center the nodes about 0 and thus add a space of half a grid constant between the outermost nodes and the boundary of the extent. If you want three nodes at -1,0,1 you thus have to set the extent to 3, i.e., stretching from -1.5 to 1.5. The main reason for this change was that topology always added this padding silently when you used periodic boundary conditions (otherwise, neurons are the left and right edge would have been in identical locations, not what one wants).
• Functions computing connection probabilities, weights and delays as functions of distance between source and target nodes now handle periodic boudary conditions correctly.
• Masks with a diameter larger than the diameter of the layer they are applied to are now prohibited by default. This avoids multiple connections when masks overwrap.

Documentation

• Updated the PyNEST documentation on the NEST homepage to reflect the current API.
• Updated all pages related to the helpdesk. This includes a face-lift for all pages as well as the addition of letter links to the alphabetical index.
• Updated the SLI tutorial on the homepage to reflect the current API.
• Updated the example networks on the homepage and added the resulting plots.
• Updated and extended the role and seeding of random number generators in NEST. See Random numbers in NEST.
• Added example for repeated stimulation using 'origin': pynest/examples/repeated_stimulation.py
• Added example of how to record data from multimeter to file: pynest/examples/multimeter.py
• Added documentation for using NEST with MUSIC.

Bug fixes

• Setting impossible values for number of virtual processes and threads is now prohibited.
• step_current_generator now keeps its last value to the end of simulation.
• Negative rates are now prohibited in poisson_generator_ps.
• Fixed several possible crashes in ht_neuron.
• Fixed static destruction problem in iaf_cond_alpha_mc.

Versions earlier than 2.0 RC1

Please refer to Software:Release_Notes_1.9.