Sutton, R., A.G., Barto, C., Anderson,1983, Neuron-like Adaptive Elements that can solve Difficult Learning Control Problems, IEEE Trans., SMC-13, 5, pp.834846.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....136 B.6 The GUI of the genome editor. List of Tables 6. 1 Features of different cell types building the artificial tissue: 1] participates in diffusion, 2] can be stimulated, 3] behavior depends on gene expression, 4] can have axons, 5] can have dendrites, [6] produces a diffusible cell type protein. 53 6.2 Condition atoms with veto power. Here, CTPx stands for the cell type protein CTPx, while [PTx] is the current local concentration of protein PTx. CTPx can be replaced by any kind of cell type protein, whereas as PTx can be any type of ....

....by a teacher telling or showing to the individual what is right and wrong, or simply physical stimulation like pain or taste (e.g. sweet nutrients) replaces the part of the teacher. The weightupdating rule for such an ANN architecture incorporates (e.g. as learning rate) the reinforcement signal [6, 7]. Unsupervised Learning Another possibility for net learning is unsupervised learning, when synapses between neurons compete with one another for their strength. Some ANN models strengthen all those synapses at or around the neuron which was stimulated the most after the net input, while all ....

[Article contains additional citation context not shown here]

R.S. Sutton, A.G. Barto and C.W. Anderson. Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Transactions on Systems, Man, and Cybernetics, 5:834--846, 1983.

....for each state, but are being driven into a certain direction as learning progresses. 2 Barto s approach The idea of unsupervised reinforcement learning (more specifically, learning without the existence of a set of learn samples) was first proposed by Barto, Sutton and Anderson (see [BA83]) They use an Associative Search Element (ASE) which determines the optimal mapping between states s, current reinforcement r and actions y using a probabilistic search, and an Adaptive Critic Element (ACE) for the prediction of future reinforcement. In control applications, a typical choice ....

....algorithm the user likes. In this assignment a modified version of the pole.c algorithm was used to control the racing car on the track. This program was extended in such a way that it uses neuro control with reinforcement. The pole.c program uses the boxes approach, as described by Sutton et al. [BA83]. Here we used the general approach as sketched in section 2, this means that the p(t) is no longer a single weight but the (weighted) sum of the weights of the critic. Our aim in this simulation was twofold. First keep the car in the middle of the track at a constant speed. Second, if we have a ....

Sutton, R. Barto, A. and Anderson, C. Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Transactions on Systems, Man and Cybernetics, SMC-13:834--846, september 1983.

....with 16 robots, all of which are completely programmed; i.e. there is no neural learning. RARS also comes with different tracks. We chose the default track, trackfil.trk, to do most our simulations. 2. 2 pole.c As an example of reinforcement learning we were given a program pole.c by Barto et al. [BA83]. In this program a small cart is controlled. The cart can drive in only two directions: left or right. A pole is attached to the cart so that it can rotate along the same axis as the cart moves (fig. 1 (a) The goal is to balance the pole by driving to the left or right. When the pole falls ....

Sutton, R. Barto, A. and Anderson, C. Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Transactions on Systems, Man and Cybernetics, SMC-13:834--846, september 1983.

No context found.

Sutton, R., A.G., Barto, C., Anderson,1983, Neuron-like Adaptive Elements that can solve Difficult Learning Control Problems, IEEE Trans., SMC-13, 5, pp.834846.

No context found.

Barton, A.G., Sutton, R.S., & Anderson, C.W. (1983). Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Transactions on Systems, Man, and Cybernetics, SMC-13, 834--846.

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