multigeom.fit#

Fitting to the paper’s recordings (extra [paper]: h5py, plenoptic).

* `train_encoder`: the paper's recipe for fitting an encoder to the macaque recordings.
  It holds the TVSD data around `NeuralEncoder.fit_readout`, which is the one place
  where a readout is fitted;
* the TVSD data loaders (`tvsd`), the observation-noise calibration and the backbone
  layer sweep.

Not needed to use an encoder that is already fitted.

Modules

`calibrate_sigma`	Calibrate the observation-noise scale σ of the Gaussian observation model
`layer_sweep`	Truncation-layer sweep for the frozen-backbone encoders (paper App.
`train_encoder`	The paper's recipe for fitting an encoder to macaque recordings: load the 50 most reliable channels of V1 or V4 from the THINGS Ventral-stream Spiking Dataset (Papale et al.), build a reference encoder (Inception, ResNet-50 or ON-OFF backbone, Gaussian readout), fit its readout and score it on the test images.
`tvsd`	Data loaders for the THINGS Ventral-Stream Spiking Dataset (TVSD) of Papale et al. (extra [paper]: needs h5py).