A Growing Neural Gas Network Learns Topologies

By Bernd Fritzke. 1994

How researchers see this paper

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

In Model Architectures

Citations in the corpus

Progressive Growing of GANs for Improved Quality, Stability, and Variation (2017, background): “In contrast to early work on adaptively growing networks, e.g., growing neural gas (Fritzke, 1995) and neuro evolution of augmenting topologies (Stanley & Miikkulainen, 2002) that grow networks greedily, we simply defer the introduction of pre-conﬁgured layers.”
Lifelong Learning of Spatiotemporal Representations With Dual-Memory Recurrent Self-Organization (2018, background): “Different models of neural network self-organization have been proposed that resemble the dynamics of Hebbian learning and plasticity (Fritzke, 1995; Kohonen, 1995; Marsland et al., 2002), with neural map organization resulting from unsupervised statistical learning.”
An Incremental Self-Organizing Architecture for Sensorimotor Learning and Prediction (2017, methodology): “algorithm [49], another widely used growing self-organizing neural network.”
Lifelong learning of human actions with deep neural network self-organization (2017, methodology): “In growing self-organizing networks such as the GNG (Fritzke, 1995) and the GWR (Marsland et al., 2002), available resources to allocate new information can be created in terms of new neurons and network connectivity driven by the input distribution.”
Few-Shot Lifelong Learning (2021, methodology): “2020), utilizes a neural gas network (Thomas and Klaus 1991; Fritzke 1995) to”
Remembering for the Right Reasons: Explanations Reduce Catastrophic Forgetting (2020, positive): “Table 2 shows the target layer name and (2020) used a neural gas network (Martinetz et al., 1991; Fritzke et al., 1995) which can learn and preserve the topology of the feature manifold formed by different classes and we have followed their experimental protocol for our CUB200 experiment by using…”
Few-Shot Class-Incremental Learning (2020, methodology): “The matching process divides F into disjoint subregions, where the centroid vector m j encodes the region Noting that some variants of NG [8, 31] use different approaches to construct NG incrementally.”
A Comprehensive Study of Class Incremental Learning Algorithms for Visual Tasks (2020, background): “NG and SOM are growing networks that were widely used for incremental semi-supervised clustering [19], multiclass online classification problems [5], and online semisupervised vector quantization learning [66].”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, methodology): “As the history grows, to find an M with small bitsM + bitsH , step 4 uses sequential network construction: it regularly changes M ’s size by adding or pruning units and connections [111, 112, 8, 168, 59, 107, 304, 176, 141, 92, 143, 204, 53, 296, 106, 31, 57, 185, 283].”
Self-Delimiting Neural Networks (2012, neutral): “This type of SLIM program-directed NN growth is quite different from previous popular NN growth strategies, e.g., [12].”
Emergence of multimodal action representations from neural network self-organization (2017, methodology): “Since real-world events unfold at multiple spatial and temporal scales, artificial neurocognitive architectures should account for the efficient processing and integration of spatiotemporal stimuli with different levels of complexity (Fonlupt, 2003; Hasson, Yang, Vallines, Heeger, & Rubin, 2008; Ler”
Self-organizing neural integration of pose-motion features for human action recognition (2015, background): “The Growing Neural Gas (GNG) proposed by Fritzke (1995) has the ability to add new neurons to an initially small network by evaluating local statistical measures on the basis of previous adaptations, and to create and remove connections between existing neurons.”
A review of algorithms for filtering the 3D point cloud (2017, unknown): “Orts-Escolano et al. [24] first used a 3D filtering and downsampling technique based on Growing Neural Gas (GNG) [25–27] network.”
Clustering: A neural network approach (2010, unknown): “, 1991a), the growing neural gas (GNG) (Fritzke, 1995a), and the FOSART (Baraldi & Parmiggiani, 1997) are surveyed and compared in Baraldi and Blonda (1999).”
A survey of dimensionality reduction techniques (2014, methodology): “Neural Gas networks [Martinetz1991, Martinetz1993, Fritzke1995] is an approach similar to the standard SOM, only that the neighborhood topology is automatically learnt from the data.”
Dynamic self-organising map (2011, methodology): “Vector quantisation can be achieved using several methods such as variations of the k -means method [1], Linde–Buzo–Gray (LBG) algorithm [2] or neural network models such as the self-30 organising map (SOM) [3], neural gas (NG) [4] and growing neural gas (GNG) [5].”
Essentials of General Intelligence: The Direct Path to Artificial General Intelligence (2007, methodology): “Growing Neural Gas, [ 10 ]), unsupervised and self-supervised learning, perceptual learning [13], and fuzzy logic [18].”