J. Vesanto, "Using the SOM and Local Models in Time-Series Prediction ", in Proc. of WSOM'97, Espoo (Finland), pp. 209-214, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....a metric based approach. Hence it can be applied directly to structural data if data are represented such that an appropriate metric for the respective structural data is defined and a notion of how to adapt within the space given by structural data can be found. This has been proposed e.g. in [16, 26, 43]. Various approaches alternatively enlarge SOM by recurrent dynamics such as leaky integrators or more general recurrent connections which allow the recursive processing of sequences. Examples are the temporal Kohonen map (TKM) 7] the recursive SOM (RecSOM) 45, 46, 47] or the approaches ....

J. Vesanto. Using the SOM and local models in time-series prediction. In: Proc. Workshop on SelfOrganizing Maps 1997.

....they are limited due to exponential loss of context. Thus far, a number of methods have been proposed for representing temporal information with the Self Organizing Map (SOM) with various successes. This includes both distributed and local types of representations. These methods use time delays [6, 11], recurrent connections [3, 4, 8] leaky integrators [1] or sometimes combine several of those principles [5, 9] However, it is not clear in what sense these models correctly extend the SOM properties to time. In general, classical notions like vector quantization and quantization error are not ....

J. Vesanto. Using the som and local models in time-series prediction. In Proceedings of WSOM'97, pages 209-214, 1997.

....and the computational complexity of its construction is of the order N 3 . Hence for a large enough N the cost of constructing a GP will be quite significant. The use of multiple local metamodels is expected to alleviate this burden. This can be achieved by adopting the strategy suggested in [18], where a SOM is used to partition the data space for constructing interpolating models. Acknowledgments This work was supported under EPSRC grant no GR L04733. ....

J. Vesanto, "Using the SOM and Local Models in Time-Series Prediction," in Proceedings

....[13] RNNs with 10 adaptive delayed connections trained with BPTT combined with a constructive algorithm. BGALR [14] A genetic algorithm with adaptable input time window size (Breeder Genetic Algorithm with Line Recombination) EPNet [15] Evolved neural nets (Evolvable Programming Net) SOM [16]: A Self organizing map. Neural Gas [17] The Neural Gas algorithm for a Vector Quantization approach. AMB [18] An improved memory based regression (MB) method [19] that uses an adaptive approach to automatically select the number of regressors (AMB) The results from these approaches are found ....

J. Vesanto, \Using the SOM and local models in time-series prediction," in Proceedings of WSOM'97, Workshop on Self-Organizing Maps, Espoo, Finland, June 4-6, pp. 209-214, Espoo, Finland: Helsinki University of Technology, Neural Networks Research Centre, 1997.

....approaches, we did not consider works where noise was added to the task or where training conditions were very di erent from ours. When not speci cally mentioned, an input time window with time delays t, t 6, t 12 and t 18 or larger was used. The di erent approaches are outlined in Table 7.1. Vesanto (1997) o ers the best result to date, according to our knowledge, with a Self Organizing Map (SOM) approach. The SOM parameters given in Table 7.2 refers to the prototype vectors of the map. The results from these approaches are found in Table 7.2. We re calculated the results for R. Bone et al. ....

....with BPTT combined with a constructive algorithm. BGALR Falco, Iazzetta, Natale, and Tarantino (1998 ) A genetic algorithm with adaptable input time window size (Breeder Genetic Algorithm with Line Recombination) EPNet Yao and Liu (1997) Evolved neural nets (Evolvable Programming Net) SOM Vesanto (1997) A Self organizing map. Neural Gas Martinez, Berkovich, and Schulten (1993) The Neural Gas algorithm for a Vector Quantization approach. AMB Bersini, Birattari, and Bontempi (1998) An improved memory based regression (MB) method (Platt, 1991) that uses an adaptive approach to automatically select ....

[Article contains additional citation context not shown here]

Vesanto, J. (1997). Using the SOM and local models in time-series prediction. In Proceedings of WSOM'97, Workshop on Self-Organizing Maps, Espoo, Finland, June 4-6 (pp. 209-214). Espoo, Finland: Helsinki University of Technology, Neural Networks Research Centre.

....provide an effective and accurate approximation, while high order polynomials in higher dimensions are not significantly better than the first order ones. For quantization of the state space, in most of the neural networks papers, the self organizing map (SOM) 5] has been used [6] 7] [8]. Apart from quantization, the SOM attempts to represent the topology of the approximated data manifold, so it can serve as a visualization tool as well. On the other hand, predefined map structure is a disturbing factor in quantization if the map dimension does not coincide with the intrinsic ....

....iterative prediction based on x(t 6) is usually applied. To make comparisons compatible, we focus only on neural networks trained on Mackey Glass timeseries with ffi = 17. Among the approaches based on self organized quantization of the state space and local linear modelling, we include SOM SVD [8] and NGN ALLM 2 [11] methods. 2 ALLM stands for adaptive LLMs, which is an incremental adaptation of parameters derived from least squares error minimization. As a best result, the NRMSE for SOM SVD was achieved in configuration 1225 units 3000 inputs, namely f0.0048,0.022g for predicting ....

J. Vesanto. Using the SOM and local models in time-series prediction. In Proc. of the WSOM'97: Workshop on Self-Organizing Maps, pages 209--214. Espoo, Finland, 1997. 3 The values of NRMSE were estimated from the line graph, given in their paper, displaying log(NMSE).

....is carried out. The estimation of the model includes searching for the threshold values and the parameters of local AR models. In the local model approach division of the data set is usually carried out with some clustering or quantization algorithm such as k means, SelfOrganizing Map (SOM) 28] [27] or neural gas [18] Another approach using self organizing capability and local models is presented in [14] After clustering the data, local models for the generated local data sets are estimated. The problem of data division for achieving the best prediction accuracy remains in a general case ....

J. Vesanto. Using the SOM and local models in time-series prediction. In Proc. of Workshop on SelfOrganizing Maps, pages 209--214. Helsinki University of Technology, 1997.

....the measured data. Local models are based on dividing the data set to smaller sets of data, each being modeled with a simple local model [9] Creation of the local data sets is usually carried out with some clustering or quantization algorithm such as k means, Self Organizing Map (SOM) 13] [12] or neural gas [7] Input to the model is usually provided by using a windowing technique to split the time series into input vectors. Typically input vectors contain past samples of the series up to certain length. In this procedure the temporal context between consecutive vectors is lost. One ....

J. Vesanto. Using the SOM and local models in time-series prediction. In Proc. of Workshop on Self-Organizing Maps, pages 209--214. Helsinki University of Technology, 1997.

....quantization performed by SOM partitions the input data space spanned by fxg into a set of regions. For each region a local model for predicting the output variable y is trained in a supervised manner. The local models can be constructed either simultaniously or after the SOM has been trained [109]. A scheme for constructing and using the local models is depicted in Fig. 4.8. The local models are formed of the data vectors falling to the Voronoi regions of the map units. If a local data set of a single map unit is considered too small, it can be augmented from the data sets of similar, ....

Juha Vesanto. Using the SOM and Local Models in Time-Series Prediction.

No context found.

J. Vesanto, "Using the SOM and Local Models in Time-Series Prediction ", in Proc. of WSOM'97, Espoo (Finland), pp. 209-214, 1997.

No context found.

Vesanto, J. (1997). Using the SOM and local models in time-series prediction. In: Proceedings of the workshop on self-organizing maps, Espoo, Finland (pp. 209--214).

No context found.

J. Vesanto, "Using the SOM and local models in time series prediction," in Proc. 1997.

No context found.

J. Vesanto, Using the SOM and local models in time series prediction, in: Proceedings of the WSOM'97: Workshop on Self-Organizing Maps, 1997.

No context found.

Vesanto, J.: Using the SOM and Local Models in Time Series Prediction, Proceedings WSOM'97: Workshop on Self-Organizing Maps, 1997.

No context found.

Juha Vesanto, Using the SOM and Local Models in TimeSeries Prediction, In Proceedings of Workshop on SelfOrganizing Maps (WSOM'97), Espoo, Finland, pp. 209-214, 1997.

No context found.

J. Vesanto. Using the SOM and local models in time-series prediction. In: Proc. Workshop on Self-Organizing Maps 1997.

No context found.

J. Vesanto, "Using the SOM and Local Models in Time-Series Prediction", In Proc. of WSOM'97, Espoo, Finland, pp. 209-214, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC