emer/axon

Issues

• Make axon Threading heuristics MPI aware

  #197 opened a year ago
  2

• add default params to layers and prjns

  #196 opened a year ago
  1

• Flexible neuron, synapse data layout

  #195 opened a year ago
  6

• US Magnitude representation

  #194 opened a year ago
  1

• Figure out more systematic treatment of negative USs in BOA, test in PVLV

  #191 opened a year ago
  2

• Add hebbian learning rule

  #188 opened 5 months ago
  0

• Issues for using MPI in GUI interactive mode

  #182 opened 2 years ago
  1

• Fix positioning of BlaNegExtD1 and BlaNegAcqD2

  #181 opened 2 years ago
  1

• Add rand.Seed to axon.Network -- use for all random calls

  #179 opened 2 years ago
  1

• Add Classes to Network.AllPrjnScales

  #178 opened 2 years ago
  1

• SynCa: test running integration loop at 2x step

  #173 opened 2 years ago
  1

• Threads.SynCa and SynCaFun no longer needed (after GPU PR merge)

  #172 opened a year ago
  3

• GPU: Make a list of neuron indexes that spiked, use that for SynCa and SendSpike

  #171 opened 2 years ago
  1

• GPU: save and re-use command buffers for RunCycleOne and RunCycles

  #169 opened 2 years ago
  1

• Data parallelism for GPU (and CPU)

  #168 opened 2 years ago
  1

• running GPU tests on CI: vulkan-swrast / lavapipe

  #159 opened 2 years ago
  1

• Matrix learning rule: should sub-pools only get credit if they were the ones that drove Gating

  #154 opened 2 years ago
  0

• Change all the `AddLayerInit` code in various *_net files to use simpler AddLayer2D or 4D

  #153 opened 2 years ago
  1

• Add ACh modulation to STN

  #150 opened a year ago
  1

• Use functional names for layers, with biological mapping

  #149 opened 2 years ago
  0

• Tests to add

  #148 opened 2 years ago
  1

• add convenient StopOnErr mech to looper, use in boa

  #145 opened 2 years ago
  1

• Get rid of all the AxonLayer, AxonPrjn interfaces and just use *Layer and *Prjn

  #144 opened 2 years ago
  1

• Target layer with nothing in plus phase is getting broad flat activity

  #142 opened 2 years ago
  1

• NonDefault params not working

  #141 opened 2 years ago
  2

• boa: didGate is not correct

  #138 opened 2 years ago
  1

• add PTLayer PostBuildDefaults method to set Thal -> PT as ModulatoryG

  #137 opened 2 years ago
  1

• for examples/pcore test case: use RW DA to prevent overlearning

  #132 opened 2 years ago
  0

• 2 Questions

  #122 opened 2 years ago
  4

• Add a way to specify the random seed

  #115 opened 2 years ago
  2

• Adding checkpointing for storing Axon Networks to disk

  #114 opened a year ago
  2

• Memory bandwidth benchmark

  #82 opened 2 years ago
  1

• SWt Limit .2 .8 causes initial weights to go in opposite direction if Mean > .5

  #54 opened 2 years ago
  2

• remove Inhib.Bg

  #51 opened 2 years ago
  0

• learning dynamics depend on `NeurCa.SpikeG` in ways that are not compensated for by lrate

  #48 opened 2 years ago
  0

• DWtSynSpkTheta uses `KinaseCa.UpdtThr` and NeurSpkTheta uses `XCal.LrnThr`

  #47 opened 2 years ago
  0

• Add FFT analysis pathway in estats etc for looking at oscillations in the models.

  #39 opened 2 years ago
  1

• remove examples/ra25example_world

  #38 opened 2 years ago
  0

• Organize params in ra25 -- minimal base with all defaults..

  #37 opened 2 years ago
  1

• Use faster form of PCA and simpler stats like variance to track representational health

  #36 opened 2 years ago
  1

• optimize threading by re-aligning steps, do all in one pass per layer

  #35 opened 2 years ago
  1

• Need A-type K channels to prevent over-activation in VmDend / NMDA

  #31 opened 3 years ago
  2

• can we replace SS with just an S with a longer time constant?

  #29 opened 3 years ago
  1

• issues for integrating synaptic inputs: max driver, per synapse..

  #28 opened 2 years ago
  3

• Check all models input-layer gi values and max activations

  #27 opened 2 years ago
  0

• Double check minus-phase plus phase ActInt clearing

  #26 opened 3 years ago
  1

• Don't learn for synapses with weight failure -- add option

  #25 opened 2 years ago
  2

• base send failures on SWt not Wt

  #24 opened 2 years ago
  2

• separate predictive learning loops from attentional ones

  #23 opened 3 years ago
  1

• Add a mechanism to allow spectrum of time constants in 6CT neurons

  #22 opened 3 years ago
  0