0.4.0

New features

• On chip building of synaptic matrices - If NativeRNG is used as rng for distributions and probabalistic connectors they will be calculated on SpiNNaker rather than the host machine, saving huge amounts of load time
• Support for multiplicative STDP weight dependence
• Code likely to prove common to other C++ based SpiNNaker projects moved out into rig_cpp_common project

Improvements

• Better performance estimation - Performance models of neurons and synapses are more accurate and these are used to build more accurate models of memory and CPU requirements of models.
• Better partitioning driven by improved performance models - No need to provide partitioning hints
• Automatic (very simple) number of boards calculation for spalloc

0.3.1

Bug fixes related to reading back synaptic matrices with delay extension

0.3.0

Changes:

• Massively reduced memory footprint by (ab)using PyNN MPI support to only build a submatrix at a time
• Removed restriction that synapse processors must be 'wider' than neuron processors
• Starting adding automated tests - heuristics for PyNN connectors, lazy param mapping system and some utils now with tests
• Some new lazy param map primitives for probabilistic models

0.2.4

Fixed typo in setup.py

0.2.3

Bug fixes relating to 'fast' Poisson spike sources (above 200Hz)

0.2.2

Updated requirements:

• To version of spalloc that fixes bug which causes pynn_spinnaker to hang on exit if an exception was thrown
• To released version of Rig that fixes issues with fixed-point conversion

0.2.1

Now was extra spalloc, just replace the spinnaker_hostname setup argument with:

import pynn_spinnaker as sim

sim.setup(timestep=1.0, spalloc_num_board=1)

0.2.0

Changes:

• Plugin system - Allows external modules to be used to supply extra neuron models etc
• Improved accuracy of exponential synapse decay
• Support for synapse models with signed weights
• Flushing mechanism - Prevents back-propagating spikes being lost in plastic networks
• Analogue recording supports sampling intervals - Saves lots of memory/time!
• ExtendedPlasticSynapticMatrix region lets you build learning rules with extra synaptic state
• Framework for intrinsic plasticity
• Lots of bug fixes

0.1

First release

• Support for IF_curr_exp and IF_cond_exp neuron models
• Support for StaticSynapse and STDPMechanism synapse models
• Support for SpikePairRule and Vogels2011Rule STDP timing dependences
• Support for AdditiveWeightDependence STDP weight dependences