Pooled texture model

docs/tutorials/models/pooled_texture_model.nblink

@@ -0,0 +1,3 @@
+{
+	"path": "../../../examples/pooled_texture_model.ipynb"
+}

src/plenoptic/simulate/models/__init__.py

@@ -1,3 +1,4 @@
 from .frontend import *
 from .naive import *
 from .portilla_simoncelli import PortillaSimoncelli
+from .portilla_simoncelli_masked import PortillaSimoncelliMasked

src/plenoptic/simulate/models/portilla_simoncelli.py

@@ -359,8 +359,7 @@ def forward(
 
         # Compute the central autocorrelation of the coefficient magnitudes. This is a
         # tensor of shape: (batch, channel, spatial_corr_width, spatial_corr_width,
-        # n_orientations, n_scales). var_mags is a tensor of shape (batch, channel,
-        # n_orientations, n_scales)
+        # n_orientations, n_scales).
         autocorr_mags, mags_var = self._compute_autocorr(mag_pyr_coeffs)
         # mags_var is the variance of the magnitude coefficients at each scale (it's an
         # intermediary of the computation of the auto-correlations). We take the square
@@ -892,8 +891,8 @@ def plot_representation(
             self,
             data: Tensor,
             ax: Optional[plt.Axes] = None,
-            figsize: Tuple[float, float] = (15, 15),
-            ylim: Optional[Union[Tuple[float, float], Literal[False]]] = None,
+            figsize: Tuple[float, float] = (15, 5),
+            ylim: Optional[Union[Tuple[float, float], Literal[False]]] = False,
             batch_idx: int = 0,
             title: Optional[str] = None,
     ) -> Tuple[plt.Figure, List[plt.Axes]]:
@@ -1113,11 +1112,7 @@ def update_plot(
         rep = {k: v[0, 0] for k, v in self.convert_to_dict(data).items()}
         rep = self._representation_for_plotting(rep)
         for ax, d in zip(axes, rep.values()):
-            if isinstance(d, dict):
-                vals = np.array([dd.detach() for dd in d.values()])
-            else:
-                vals = d.flatten().detach().numpy()
-
+            vals = to_numpy(d.flatten())
             sc = update_stem(ax.containers[0], vals)
             stem_artists.extend([sc.markerline, sc.stemlines])
         return stem_artists