m8ta
You are not authenticated, login.
text: sort by
tags: modified
type: chronology
{1568}
hide / / print
ref: -2021 tags: burst bio plausible gradient learning credit assignment richards apical dendrites date: 05-05-2022 15:44 gmt revision:2 [1] [0] [head]

Burst-dependent synaptic plasticity can coordinate learning in hierarchical circuits

  • Roughly, single-events indicate the normal feature responses of neurons, while multiple-spike bursts indicate error signals.
  • Bursts are triggered by depolarizing currents to the apical dendrites, which can be uncoupled from bottom-up event rate, which arises from perisomatic inputs / basal dendrites.
  • The fact that the two are imperfectly multiplexed is OK, as in backprop the magnitude of the error signal is modulated by the activity of the feature detector.
  • "For credit assignment in hierarchical networks, connections should obey four constraints:
    • Feedback must steer the magnitude and sign of plasticity
    • Feedback signals from higher-order areas must be multipleed with feedforward signals from lower-order areas so that credit assignment can percolate down the hierarch with minimal effect on sensory information
    • There should be some form of alignment between feedforward and feedback connections
    • Integration of credit-carrying signals should be nearly linear to avoid saturation
      • Seems it's easy to saturate the burst probability within a window of background event rate, e.g. the window is all bursts to no bursts.
  • Perisomatic inputs were short-term depressing, whereas apical dendrite synapses were short-term facilitating.
    • This is a form of filtering on burst rates? E.g. the propagate better down than up?
  • They experiment with a series of models, one for solving the XOR task, and subsequent for MNIST and CIFAR.
  • The later, larger models are mean-field models, rather than biophysical neuron models, and have a few extra features:
    • Interneurons, presumably SOM neurons, are used to keep bursting within a linear regime via a 'simple' (supplementary) learning rule.
    • Feedback alignment occurs by adjusting both the feedforward and feedback weights with the same propagated error signal + weight decay.
  • The credit assignment problem, or in the case of unsupervised learning, the coordination problem, is very real: how do you change a middle-feature to improve representations in higher (and lower) levels of the hierarchy?
    • They mention that using REINFORCE on the same network was unable to find a solution.
    • Put another way: usually you need to coordinate the weight changes in a network; changing weights individually based on a global error signal (or objective function) does not readily work...
      • Though evolution seems to be quite productive at getting the settings of (very) large sets of interdependent coefficients all to be 'correct' and (sometimes) beautiful.
      • How? Why? Friston's free energy principle? Lol.

{159}
hide / / print
ref: Wichmann-2011.12 tags: DBS STN basal ganglia bursts oscillation review wichmann beta date: 02-22-2012 17:05 gmt revision:13 [12] [11] [10] [9] [8] [7] [head]

PMID-21723919[0] Pathological basal ganglia activity in movement disorders.

  • The paradigm has shifted: initial idea was that firing rates changed,
  • later in detailed description of basal ganglia firing rate changes:
    • burst patterns and oscillations
  • 6-OHDA murines + MPTP monkey models so essential yada yada.
  • intraoperative microelectrode recordings yada yada.
  • Nice figure:
    • Black = inhibitory; gray = excitatory. From Galvan and Wichmann 2008.
    • note differences between D2 and D1.
  • Recall corticostriatal fibers are often (50%) collaterals from corticospinal axons.
  • Corticostriatal pathway separate from cortico-subthalamic pathway, so the two get different signals. (Parent and Parent 2006).
    • Few collaterals, and of those axons go to red nucleus and cerebral peduncle -- not pyramids.
  • Indirect (GPe, STN targets) and direct (GPi/SNr) striatal projections generally, but not completely, seem separate.
  • VA = ventroanterior; VL = ventrolateral thalamus.
  • Collaterals from GPi/SNr reach the intralaminar thalamic nuclei: the CM (centromedian) and the PF (parafascicular) nuclei.
  • One of the important additional function of the intralaminar thalamic nuclei is to provide saliency information to the striatum during procedural learning (Kimura et al 2004; Minamimoto et al 2009).
  • There is a considerable body of evidence that the absence of dopaminergic transmission may trigger changes in the density and morphology of dendritic spines on striatal projection neurons.
    • Thereby influencing corticostriatal transmission.
    • This is consistent with the progressive nature of the disease.
  • Serotonin and acetylcholine also involved in striatum, but their role in PD less well characterized.
  • Tremor and dystonia possibly due to afferents from the deep cerebellar nuclei and efferents to the cerebellar cortex.
  • Rate model failures:
    • thalamotomy procedures did not result in worsening of parkinsonism.
    • GPi lesions produced bradykinesia in normal monkeys (despite the GABA output!)
    • GPe lesions do not produce parkinsonism.
    • not all studies report changes in FR in GPi/GPe.
    • A significant factor interfering with the assessment of FR changes in PD patients is that its dependent on the state of arousal of the patients.
  • Burstiness: Increased burstiness (Fig. 2A) has emerged as one of the most reliable abnormalities of neuronal firing in the basal ganglia in parkinsonism, as shown in dopamine-depleted monkeys and in patients with PD
  • Oscillations: much in the beta band (10-35 Hz) throughout extrastriatal BG.
old redirect: see [1]
  • LFP power:
  • Brown is the purveyor of the high kinetic / low akinetic hypothesis (2003, 2005).
  • Oscillations do not occur in acute dopamine depletion.
  • GABA receptor blockade in GPe results in dyskinesias.
  • STN inactivation results in ballismus, as noted elsewhere.
  • GPi lesioning is clinically used to abolish dyskinesias in patients with treatment-resistant hyperkinetic movements.

____References____

[0] Wichmann T, Dostrovsky JO, Pathological basal ganglia activity in movement disorders.Neuroscience 198no Issue 232-44 (2011 Dec 15)
[1] Rodriguez-Oroz MC, Rodriguez M, Guridi J, Mewes K, Chockkman V, Vitek J, DeLong MR, Obeso JA, The subthalamic nucleus in Parkinson's disease: somatotopic organization and physiological characteristics.Brain 124:Pt 9, 1777-90 (2001 Sep)

{1094}
hide / / print
ref: Chan-2011.01 tags: GPi DBS burst oscillations date: 01-26-2012 17:29 gmt revision:1 [0] [head]

PMID-20727974[0] Bursts and oscillations as independent properties of neural activity in the parkinsonian globus pallidus internus.

  • 132 GPi cells in 14 Parkinson's disease patients
  • segregated into oscillatory and non-oscillatory cells.
    • No significant difference in burstiness between the two.
    • Except for oscillatory cells in the alpha region (8-13Hz), which were burstier than other cells.
  • conclusion: busrting and oscillations may be caused by different things.

____References____

[0] Chan V, Starr PA, Turner RS, Bursts and oscillations as independent properties of neural activity in the parkinsonian globus pallidus internus.Neurobiol Dis 41:1, 2-10 (2011 Jan)

{82}
hide / / print
ref: Wilson-2006.12 tags: parkinsons burst firing MPTP mice STN DBS date: 01-26-2012 17:25 gmt revision:6 [5] [4] [3] [2] [1] [0] [head]

PMID-16973296[0] Subthalamic nucleus neurones in slices from MPTP-lesioned mice show irregular, dopamine-reversible firing pattern changes, but without synchronous activity

  • loss of dopamine in parkinson-model rats (not MPTP!) induces synchronized low-frequency oscillatory burst-firing in subthalamic nucleus neurons
  • MPTP mice, neurons fire slower, and more irregularly
  • only dopamine varied (increased) firing rate.
  • the STN & GP are insufficient to generate the abberant firing patterns in the STN, by itself - the disease is more than just dopamine depletion.

____References____

[0] Wilson CL, Cash D, Galley K, Chapman H, Lacey MG, Stanford IM, Subthalamic nucleus neurones in slices from 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-lesioned mice show irregular, dopamine-reversible firing pattern changes, but without synchronous activity.Neuroscience 143:2, 565-72 (2006 Dec 1)