m8ta
You are not authenticated, login.
text: sort by
tags: modified
type: chronology
{1576}
hide / / print
ref: -0 tags: GFlowNet Bengio probabilty modelling reinforcement learing date: 10-29-2023 19:17 gmt revision:3 [2] [1] [0] [head]

GFlowNet Tutorial

  • It's basically like RL, only treating the reward as a scaled unnormalized probability or 'flow'.
  • Unlike RL, GFN are constructive, only add elements (actions), which means the resulting graph is either a DAG or tree. (No state aliasing)
  • Also unlike RL / REINFORCE / Actor-critic, the objective is to match forward and reverse flows, both parameterized by NNs. Hence, rather than BPTT or unrolls, information propagation is via the reverse policy model. This forward-backward difference based loss is reminiscent of self-supervised Barlow Twins, BYOL, Siamese networks, or [1][2]. Bengio even has a paper talking about it [3].
    • The fact that it works well means that it must be doing some sort of useful regularization, which is super interesting.
    • Or it just means there are N+1 ways of skinning the cat!
  • Adopting a TD(λ) TD(\lambda) approach of sampling trajectories for reward back-propagation improves convergence/generalization. Really not that different from RL..
  • At least 4 different objectives (losses):
    • Matching per-state in and out flow
    • Matching per-state forward and backward flow
    • Matching whole-trajectory forward and backward flow
    • Subsampling whole-trajectory and matching their flow.

{1238}
hide / / print
ref: -0 tags: histology immune response otto indiana electrodes gfap inflamation transparent clearing vimentin date: 04-19-2013 23:59 gmt revision:4 [3] [2] [1] [0] [head]

PMID-23428842 Chronic intracortical microelectrode arrays induce non-uniform, depth-related tissue responses.

  • Woolley AJ, Desai HA, Otto KJ.
  • One timepoint, 4 weeks.
  • Laser confocal microscopy
    • after tissue clearing (optical index of refraction matching) in a 60% sucrose solution.
  • Single-shank iridium contact silicon substrate MEA.
    • Device cut level with surface of brain after insertion.
  • Intact MEAs via device-capture histology, DHhist (Woolley et al 2011)
    • 350-450um tissue explanted with device.
    • They promote their technique.
  • Tissue surrounding microdevices exhibited two major depth-related phenomena:
    • a non-uniform microglial coating along the device length and
    • a dense mass of cells surrounding the implant in cerebral cortical layers I and II.
      • The dense mass of cells contained vimentin, a protein not typically expressed highly in CNS cells, evidence that non-CNS cells likely descended down the face of the penetrating devices from the pial surface.
        • But no Iba1 (activated microglia) per se in the tissue mass.
    • Hoe342 -- cell marker.
    • This mass was apparently consistent across animals!
    • Cells in the mass were VIM positive -- fibroblasts -- meninges?
  • low GFAP = not an astrocytic scar.
  • This study provides further evidence that a progressive invasion of non-CNS cells contributes substantially to the chronic phase of the tissue response around intracortical MEAs.
    • Again, might be from BBB distruption {1237}


This result is supported by previous papers:
  • {1193} -- microglia response not correlated to electrode failure, but correlated to ferritin immunoresponse
  • {781} -- also note that menigeal fibroblasts migrate down electrode tracts.
  • {1028} -- measured vimentin, GFAP, and ED1 (not Iba1). Found Vim+ and GFAP+, suggesting reactive astrocytes and not meningeal cells. ED1 aka CD68 is specific to macrophages and not microglia, so these may be blood-derived cells.
  • {1200} -- chronic contact with the meninges v.s intraparenchymal correlated with Vim+ encapsulation.
  • {1210} -- old paper showing the same result near surface of implant.
  • {1196} -- more against GFAP & pro BBB disruption
  • {1204} -- GFAP uncorrelated (!) with NeuN intensity
  • {307} -- all initial tests of utah arrays showed fibrous encapsulation; one array was completely explanted. This is why now they put gore-tex over the implant -- to prevent fibroblast migration (i guess).