doc upgrade, now integrates harmonium properly

Author: ago109

docs/demonstrations/demo6_boltzmann.md

@@ -337,8 +337,8 @@ in which they were produced by the Markov chain constructed above.
 
 We finally take the Harmonium that we have constructed above and fit it to
 some MNIST digits (the same dataset we used in Demonstration \#1). Specifically,
-we will leverage the `Harmonium` model in the Model Museum as it implements the
-above core components/functions internally. In the  
+we will leverage the [Harmonium](../museum/harmonium.md), model in the Model Museum
+as it implements the above core components/functions internally. In the  
 script `sim_train.py`, you will find a general simulated training loop similar to
 what we have developed in previous demonstrations that will fit our Harmonium
 to the MNIST database by cycling through it several times, saving the final

docs/index.rst

@@ -34,6 +34,7 @@ Welcome to ngc-learn's documentation!
    museum/gncn_pdh
    museum/gncn_t1_ffm
    museum/gncn_t1_sc
+   museum/harmonium
 
 .. toctree::
    :maxdepth: 1

docs/museum/gncn_pdh.md

@@ -15,7 +15,7 @@ formed by this NGC system.
     :noindex:
   .. automethod:: settle
     :noindex:
-  .. automethod:: update
+  .. automethod:: calc_updates
     :noindex:
   .. automethod:: clear
     :noindex:

docs/museum/gncn_t1.md

@@ -15,7 +15,7 @@ formed by this NGC system.
     :noindex:
   .. automethod:: settle
     :noindex:
-  .. automethod:: update
+  .. automethod:: calc_updates
     :noindex:
   .. automethod:: clear
     :noindex:

docs/museum/gncn_t1_ffm.md

@@ -15,7 +15,7 @@ vector outputs. Note that "FFM" denotes "feedforward mapping".
     :noindex:
   .. automethod:: settle
     :noindex:
-  .. automethod:: update
+  .. automethod:: calc_updates
     :noindex:
   .. automethod:: clear
     :noindex:

docs/museum/gncn_t1_sc.md

@@ -15,7 +15,7 @@ formed by this NGC system.
     :noindex:
   .. automethod:: settle
     :noindex:
-  .. automethod:: update
+  .. automethod:: calc_updates
     :noindex:
   .. automethod:: clear
     :noindex:

docs/museum/gncn_t1_sigma.md

@@ -15,7 +15,7 @@ formed by this NGC system.
     :noindex:
   .. automethod:: settle
     :noindex:
-  .. automethod:: update
+  .. automethod:: calc_updates
     :noindex:
   .. automethod:: clear
     :noindex:

docs/museum/harmonium.md

@@ -0,0 +1,28 @@
+# Harmonium (Smolensky, 1986)
+
+This circuit implements the Harmonium model proposed in (Smolensky, 1986) [1].
+Specifically, this model is **unsupervised** and can be used to process sensory
+pattern (row) vector(s) `x` to infer internal latent states. This class offers,
+beyond settling and update routines through Contrastive Divergence (Hinton 1999) [2], 
+a block Gibbs sampling function to generate a chain of synthesized patterns.
+
+```{eval-rst}
+.. autoclass:: ngclearn.museum.harmonium.Harmonium
+  :noindex:
+
+  .. automethod:: sample
+    :noindex:
+  .. automethod:: settle
+    :noindex:
+  .. automethod:: calc_updates
+    :noindex:
+  .. automethod:: clear
+    :noindex:
+```
+
+**References:** <br>
+[1] Smolensky, P. "Information Processing in Dynamical Systems: Foundations of
+Harmony Theory." Parallel distributed processing: explorations in the
+microstructure of cognition 1 (1986).
+[2] Hinton, Geoffrey E. "Training products of experts by maximizing contrastive
+likelihood." Technical Report, Gatsby computational neuroscience unit (1999).

docs/museum/model_museum.md

@@ -45,7 +45,7 @@ Current models that we have implemented in the Model Museum so far include:
 3. GNCN-PDH - the model proposed in (Ororbia & Kifer, 2022) [1]
 4. GNCN-t1-FFM - the model developed in (Whittington & Bogacz, 2017) [4]
 5. GNCN-t1-SC - the model proposed in (Olshausen & Field, 1996) [5]
-6. Harmonium - the model developed in (Hinton 1999) [6]
+6. Harmonium - the model developed in (Smolensky, 1986; Hinton 1999) [6] [7]
 
 (If there is a model you think should be exhibited/integrated into the Model
 Museum, and/or would like to contribute, please write us at [email protected] or
@@ -66,4 +66,7 @@ synaptic plasticity." Neural computation 29.5 (2017): 1229-1262. <br>
 [5] Olshausen, B., Field, D. Emergence of simple-cell receptive field properties
 by learning a sparse code for natural images. Nature 381, 607–609 (1996). <br>
 [6] Hinton, Geoffrey E. "Training products of experts by maximizing contrastive
-likelihood." Technical Report, Gatsby computational neuroscience unit (1999).
+likelihood." Technical Report, Gatsby computational neuroscience unit (1999). <br>
+[7] Smolensky, P. "Information Processing in Dynamical Systems: Foundations of
+Harmony Theory." Parallel distributed processing: explorations in the
+microstructure of cognition 1 (1986).