Merge SOM for temporal data

By Marc Strickert, Barbara Hammer. 2005

132 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Control and Systems Engineering, Signal Processing

How researchers see this paper

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

In Artificial Intelligence

In Control and Systems Engineering

In Signal Processing

Citations in the corpus

Lifelong Learning of Spatiotemporal Representations With Dual-Memory Recurrent Self-Organization (2018, background): “Empirically selecting a convenient number of neurons can be tedious for networks with recurrent dynamics, especially when dealing with non-stationary input distributions (Strickert and Hammer, 2005).”
Lifelong learning of human actions with deep neural network self-organization (2017, methodology): “We equip the GWR network with recurrent neurons following previous formulations of context learning (Strickert & Hammer, 2005), so that the best-matching neuron wb for the input x(t) is computed as follows: b = argmin i {di}, (7) di = αw · ∥x(t) − wi∥2 + K∑ k=1 αk · ∥Ck(t) − cki ∥ 2 , (8) Ck(t) = β…”
Emergence of multimodal action representations from neural network self-organization (2017, background): “It has been shown that lexical features can be learned using recursive self-organizing architectures (Strickert & Hammer, 2005), obtaining action ultimodal action representations from neural network self-orga016/j.cogsys.2016.08.002 word representations from a phonemic representation of recognized…”
Self-organizing neural integration of pose-motion features for human action recognition (2015, methodology): “…of self-organizing networks have been extensively investigated that extend the feed-forward learning mechanism with context neurons for referring to past activations, thereby allowing the processing of sequences and structured data (e.g., see Strickert and Hammer, 2005 for a recursive SOM model).”
Human motion assessment in real time using recurrent self-organization (2016, methodology): “We compared the performance of our MGWR on a time series analysis task with other two well-established models of recursive self-organization: Merge Neural Gas (MNG) [15] and Merge Growing Neural Gas (MGNG) [17].”
Nonlinear Time Series Analysis by Using Gamma Growing Neural Gas (2012, unknown): “In Merge SOM (MSOM) [4], the context is described by a linear combination of the weight and the context of the last winner neuron.”
Lifelong Learning of Action Representations with Deep Neural Self-Organization (2017, methodology): “We introduce a temporal GWR network that equips each neuron with an arbitrary number of context descriptor (Strickert and Hammer 2005) to increase the memory depth and temporal resolution in the spirit of a Gamma memory model (de Vries and Prı́ncipe 1992).”
Clustering: A neural network approach (2010, unknown): “All thesemodels introduce recurrence into the SOM, and have been reviewed and compared in Hammer, Micheli, Sperduti, and Strickert (2004) and Strickert and Hammer (2005).”
Gamma-Filter Self-Organizing Neural Networks for Time Series Analysis (2011, methodology): “Several extensions of self-organizing feature maps (SOMs) [1] for dealing with processing data sequences that are temporally or spatially connected, such as words, DNA sequences, time series, etc. [ 2 ],[3],[4] have been developed.”
Class Incremental Learning for Video Action Classification (2021, methodology): “Each neuron has a weight w i and a context descriptor c i [22, 23], which represent the weights of the current frame and the previous frames.”
Sequence learning using the adaptive suffix trie algorithm (2012, methodology): “The Merge SOM was proposed as a solution[20].”
Incremental Unsupervised Time Series Analysis Using Merge Growing Neural Gas (2009, methodology): “The Mackey Glass time series is a 1-dimensional, continuous and chaotic function defined by the differential equation dx = −0.1x(τ )+ 0.2x(τ −17) 1+x(τ −17)10 and is commonly used to evaluate the temporal quantization of recursive models [13, 1 ,24].”