doc edit

Author: ago109

docs/museum/harmonium.md

@@ -3,7 +3,7 @@
 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], 
+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}
@@ -23,6 +23,6 @@ a block Gibbs sampling function to generate a chain of synthesized patterns.
 **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).
+microstructure of cognition 1 (1986).<br> 
 [2] Hinton, Geoffrey E. "Training products of experts by maximizing contrastive
 likelihood." Technical Report, Gatsby computational neuroscience unit (1999).