Doya K (1997) Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Comp. 10:113--132.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....expensive to train than feedforward networks. Two broad classes of training algorithms are typically used: recurrent extensions of traditional feedforward learning algorithms (Pearlmutter, 1990) and evolutionary algorithms (Beer and Gallagher, 1992; Cliff, Harvey and Husbands, 1993; Nakahara and Doya, 1998; Di Paolo, 1998; Ijspeert, 2001) Both of these approaches have their advantages and disadvantages. Here we focus on a general technique for improving the performance of evolutionary searches on continuous time recurrent neural networks (CTRNNs) In a previous analysis of the general dynamical ....

Nakahara, H. and Doya, K. 1998. Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Computation 10, 113-132.

....never been observed. Such models are therefore more likely to simulate long term memory. An alternative mechanism for working memory is that information is stored by some sustained activity of neurons, which can be archived through the dynamics of recurrent neuronal network [9] Nakahara and Doya [10] have recently discussed such a network and showed that the dynamic of such a model close to a saddle node bifurcation can combine long term maintenance with quick transitions of memories. However, the storage capacities have so far not been investigated within those networks. Another very ....

H. Nakahara and K. Doja, Near-Saddle-Node Bifurcation Behavior as Dynamics in Working Memory for Goal-Directed Behavior, Neural Comp. 10, 113-132 (1998).

....plots, which run together into a single line as the output of BS passes the left edge of the fold. This can also be seen in the time plots (Figure 2B) in which BS s output turns off very slowly during the stance phase. Similar phenomena have been observed in other evolved model neural networks (Doya, 1997). Expressions that quantitatively describe the duration of this transition given the parameters of BS (Tables 1 and 2) are derived in the Appendix. For a near fold transition, the quadratic approximation tq BS BS FT BS BS Tww o o on off , predicts that the transition duration for BS ....

Doya, K. (1997) Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Comp. 10:113-132.

....Given that the autapse model is a simple and natural abstraction of recurrent networks, it is not surprising that there have been previous studies of it. These studies di er from the present model either in their lack of biophysical realism [Kamath and Keller, 1976, Cannon et al. 1983, Nakahara and Doya, 1998] or by not considering analog memory storage [Ermentrout, 1998b] 2 The conductance based model We begin by describing an autapse 1 model based on the dynamics of intrinsic and synaptic conductances [Ermentrout, 1998b] The membrane potential V of the model neuron obeys the current balance ....

Nakahara, H. and Doya, K. (1998). Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Comput, 10(1):113-32. 14

No context found.

Doya K (1997) Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Comp. 10:113--132.

No context found.

Nakahara , H., & Doya, K. (1998). Near-saddle-node bifurcation behavior as dynamics in working memory for goal-directed behavior. Neural Computation, 10, 113--132.

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