Releases: shelfwise/receptivefield
Releases · shelfwise/receptivefield
Initial Pytorch API
Release 0.4.0
Major Features And Improvements
- Added initial Pytorch API with
PytorchReceptiveField
.
Multiple feature maps
Release 0.3.0
Major Features And Improvements
- Update Keras and Tensorflow api for multiple feature maps calculation
- Change of api in compute function: removed
output_layer: str
and replaced withoutput_layers: List[str]
, which can accept a list
of names for feature maps for which we want to compute receptive fields. - Change in
plot_gradient_at
, added new parameter which control for
which feature map gradient is plotted. Same forplot_rf_grid
function. - New functions for plotting:
plot_rf_grids
andplot_gradients_at
. - Improved tensorflow API, now one does have to initialize model with
InteractiveSession, this is handled by theTFReceptiveField
. - A new API for tensorflow:
TFFeatureMapsReceptiveField
.
Beta version
First initial version of the repo