Emergence of invariant-feature detectors in the adaptive-subspace self-organizing map

By Teuvo Kohonen. 1996

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Disentangling Factors of Variation via Generative Entangling (2012, background): “The principle that invariant features can actually emerge , using only unsupervised learning, from the organization of features into subspaces was ﬁrst established in the ASSOM model (Kohonen, 1996).”
Group Equivariant Convolutional Networks (2016, methodology): “Then we take a look at functions on groups (used to model feature maps in G-CNNs) and their transformation properties.”
Deep Learning of Representations: Looking Forward (2013, background): “ameters), as in transforming auto-encoders (Hinton et al., 2011), spike-and-slab RBMs with pooled slab variables (Courville et al., 2011) and other pooling-based models that learn a feature subspace (Kohonen, 1996; Hyvarinen and Hoyer, 2000). • Ahierarchicalorganizationofexplanatoryfactors: the conc”
A Survey of Inductive Biases for Factorial Representation-Learning (2016, methodology): “Pooling can be used in an overlapping spatial grid to create a topographic map representation, as in Kohonen’s adaptive-self organizing maps [30].”
Learning Invariance from Natural Images Inspired by Observations in the Primary Visual Cortex (2012, methodology): “…such as faces on different backgrounds (Wallis & Rolls, 1997; Stringer & Rolls, 2000), more recent work is based on natural images and video sequences (Kohonen, 1996; Hyv¨arinen and Hoyer, 2000, 2001; Kayser et al., 2001; Einh¨auser et al., 2002; Hashimoto, 2003; K¨ording et al., 2004; Berkes &…”
Nonlinear Extraction of Independent Components of Natural Images Using Radial Gaussianization (2009, background): “…of local image densities arises naturally from local oriented image features that occur with arbitrary phases or angles (Zetzsche & Barth, 1990; Kohonen, 1996; Zetzsche & Krieger, 1999; Parra et al., 2001), and these concepts have been incorporated in some recent approaches for unsupervised…”
Extracting Slow Subspaces from Natural Videos Leads to Complex Cells (2001, background): “Another network for learning transformation invariant filters is the adaptivesubspace self-organizing map (ASSOM) proposed by Kohonen (1996).”
Emergence of Phase- and Shift-Invariant Features by Decomposition of Natural Images into Independent Feature Subspaces (2000, methodology): “5 Discussion An approach closely related to ours is given by the adaptive subspace self-organizing map (Kohonen, 1996), which is based on competitive learning of a invariant-feature subspace representation similar to ours.”
Learning invariant features through topographic filter maps (2009, methodology): “The idea relies on imposing sparsiﬁcation criteria on small groups of ﬁlter outputs [10, 6, 8], which can be related to the Group Lasso method for regularization [27].”
Clustering: A neural network approach (2010, unknown): “Adaptive subspace SOM (ASSOM) (Kohonen, 1996, 1997; Kohonen, Oja, Simula, Visa, & Kangas, 1996) is a modular neural network model comprising an array of topologically ordered SOM submodels.”
Essentials of the self-organizing map (2013, methodology): “A special kind of dissimilarity or distance measure is applied in an SOM that is called the Adaptive-Subspace SOM (ASSOM) , cf. Kohonen (1995, 1996, 2001) and Kohonen, Kaski, and Lappalainen (1997).”
A two-layer sparse coding model learns simple and complex cell receptive fields and topography from natural images. (2001, methodology): “The results that are closest to our model in this respect were obtained by the ASSOM model (Kohonen, 1996), but even in that model the nature of the topography was strongly in uenced by an arti cial manipulation of the input (a sampling window that moves smoothly in time).”