H. D. Block. The Perceptron: A model for brain functioning. Reviews of Modern Physics, 34: 123--135, 1962.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....discussion, we limit ourselves to neural network architectures that could feasibly be made to operate at the high speeds required for branch prediction. We consider four methods: multi layer perceptrons with back propagation, the ADALINE neuron [60] Hebb learning [23] and the Block perceptron [9]. In preliminary work, we measured the misprediction rates yielded by each method on the SPEC95 benchmarks. Hebb learning, ADALINE neurons and Block perceptrons are simple neural learning methods, in which a single neuron is used for computation and is trained with a simple algorithm. Hebb ....

....x i . x n y S S S S Sw B B B B BN w i w n Figure 5.1: Perceptron Model. 49 5.2.1 How Perceptrons Work The perceptron was introduced in 1962 [50] as a way to study brain function. We consider the simplest of many types of perceptrons [9], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function t(x 1 ; x n ) of n inputs. In our case, the x i are the bits of a global branch history shift register, and the ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....traces should be fetched from the trace cache. 3.3 Neural Learning for Dynamic Branch Prediction There are several simple neural learning methods that could potentially be used in a dynamic branch predictor. In particular, the ADALINE neuron [31] Hebb learning [11] and the Block perceptron [2] are simple methods in which a single neuron is used for computation and trained with a simple algorithm. We used the SPEC95 benchmarks to compare the accuracy of each of these methods. We also evaluated the accuracy of a more complex multi layer perceptron with back propagation [11] This ....

....products with the weights w 1 ; wn . These products are summed, along with the bias weight w 0 , to produce the output value y. 4.1 How Perceptrons Work The perceptron was introduced in 1962 [24] as a way to study brain function. We consider the simplest of many types of perceptrons [2], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function t(x 1 ; x n ) of n inputs. In our case, the x i are the bits of a global branch history shift register, and the ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....of the learning algorithm, we compare it to the performance of the best predictor in a given class. Within this framework, we introduce a new on line algorithm called second order Perceptron algorithm. This might be viewed as a variant of the classical ( rst order) Perceptron algorithm [26, 7, 24], and is similar in spirit to an algorithm introduced by Vovk [29] and further studied by Azoury and Warmuth [5] for solving linear regression problems under the square loss. Our second order Perceptron algorithm is able to exploit certain spectral properties of the data. We analyze the ....

....sequence of examples is linearly separable with margin = min t y t u x t 0 for some unit norm vector u. Then also the whitened sequence is linearly separable. In fact, hyperplane z = M u separates the whitened sequence with margin = M . By the Perceptron convergence theorem [7, 24], the number of mistakes made by the Perceptron algorithm on the whitened sequence is thus at most Mu : 1) To appreciate the potential advantage of whitening the data, notice that when the instance vectors x 1 ; x T are very correlated the quadratic form x tends to be ....

Block, H.D. (1962). The perceptron: A model for brain functioning. Reviews of Modern Physics, 34, 123-135.

....and will be described in future reports. In fact, some of our devices employ both approaches, with perpendicular gradients in temperature and catalyst composition. 3 knowledge based artificial intelligence methods. In this report we test this approach when implemented via the neural network [2] with training by back propagation method [3] Sensors The sensor arrays we use in these experiments are fabricated on commercial alumina ceramic substrates intended for hybrid circuits. The area of the substrate is compatible with a 20 pin dual inline integrated circuit. A gold electrode ....

H. D. Block, "The Perceptron: A Model for Brain Functioning", Reviews of Modern Physics, Vol. 34, No. 1, January 1962, pp. 123-35.

....discussion, we limit ourselves to neural network architectures that could feasibly be made to operate at the high speeds required for branch prediction. We consider four methods: multi layer perceptrons with back propagation, the ADALINE neuron [60] Hebb learning [23] and the Block perceptron [9]. In preliminary work, we measured the misprediction rates yielded by each method on the SPEC95 benchmarks. Hebb learning, ADALINE neurons and Block perceptrons are simple neural learning methods, in which a single neuron is used for computation and is trained with a simple algorithm. Hebb ....

....X Y Z [ a ] b b b b b c ] U d d d d d ] 1 Figure 5.1: Perceptron Model. 49 5.2.1 How Perceptrons Work The perceptron was introduced in 1962 [50] as a way to study brain function. We consider the simplest of many types of perceptrons [9], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function fg B hjikilikh V1 F inputs. In our case, the mU are the bits of a global branch history shift register, and ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....traces should be fetched from the trace cache. 3.3 Neural Learning for Dynamic Branch Prediction There are several simple neural learning methods that could potentially be used in a dynamic branch predictor. In particular, the ADALINE neuron [31] Hebb learning [11] and the Block perceptron [2] are simple methods in which a single neuron is used for computation and trained with a simple algorithm. We used the SPEC95 benchmarks to compare the accuracy of each of these methods. We also evaluated the accuracy of a more complex multi layer perceptron with back propagation [11] This ....

....with the weights 78 9.1020203.47:5 . These products are summed, along with the bias weight 7: to produce the output value . 4.1 How Perceptrons Work The perceptron was introduced in 1962 [24] as a way to study brain function. We consider the simplest of many types of perceptrons [2], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function = A BCBCBC ED of inputs. In our case, the F are the bits of a global branch history shift register, ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....a branch instruction is encountered, the address of the branch is hashed to select a neuron from the table. The selected neuron is used to predict the likely direction of the branch. Once the actual branch outcome is known, the neuron is updated with the training rule. We use the Block neuron [11], a type of perceptron [12, 13] because it provides good accuracy and lends itself to a fast and compact representation in hardware. Although perceptrons cannot learn linearly inseparable functions with 100 accuracy [14] we have found that they work well in practice for branch prediction. The ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....: m. It is useful to define the maximum achievable margin fl on a separable training set as fl = maxQ2 n fl Q = maxQ2 n min i;y2G(x i ) y 6=y i OE(x i ;y i ) DeltaQ GammaOE(x i ;y) DeltaQ jjQjj . The following theorem can then be stated: Theorem 4 (Simple modification of theorem from [Block 1962; Novikoff 1962] see also [Freund and Schapire 1999] Let f(x 1 ; y 1 ) x n ; yn )g be a sequence of examples such that 8i; 8y 2 G(x i ) jjOE(x i ; y i ) Gamma OE(x i ; y)jj R. Assume the sequence is separable, and take fl to be the maximum achievable margin on the sequence. Then the ....

....2 G(x i ) jjOE(x i ; y i ) Gamma OE(x i ; y)jj R. Assume the sequence is separable, and take fl to be the maximum achievable margin on the sequence. Then the number of mistakes made by the perceptron algorithm on this sequence is at most (R=fl) 2 . Proof: Simple modification of the proof by [Block 1962; Novikoff 1962] see also [Freund and Schapire 1999] This theorem implies that if the training sample in figure 2 is separable, and we iterate the algorithm repeatedly over the training sample (i.e. T 1) then the algorithm converges to a parameter setting that classifies the training set ....

Block, H. D. 1962. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34, 123--135.

....the proposition does not make explicit use of the generalized Blackwell condition. 4 The quasi additive algorithm In this section, we show that the quasi additive algorithm of Grove, Littlestone and Schuurmans (whose speci c instances are the p norm Perceptron [13, 15] the classical Perceptron [2, 24, 25], and the zero threshold Winnow algorithm [23] is a special case of our general decision strategy. Then, we derive performance bounds as corollaries of our Theorem 3. We recall that the quasi additive algorithm performs binary classi cation of attribute vectors x = x 1 ; xN ) 2 IR N ....

....M t = P t s=1 1l fy t 6=by t g the total number of mistakes made by the speci c quasi additive algorithm being considered. 4. 1 The p norm Perceptron As de ned in [15] the p norm Perceptron uses the polynomial potential (2) We now derive a generalization of the Perceptron convergence theorem [2, 24] (a version slightly stronger than ours was proven in [13] Fix v 0 2 IR N and 0. For an arbitrary sequence (x 1 ; y 1 ) x t ; y t ) of labeled attribute vectors, let D t = P t s=1 maxf0; y t x t v 0 g be the total deviation [9, 13, 14] of v 0 with respect to margin 0. ....

H.D. Block. The Perceptron: a model for brain functioning. Review of Modern Physics, 34:123-135, 1962.

....and Warmuth [22] in the context of (generalized) linear regression. These algorithms are called quasi additive in [16] and general additive in [22] This class includes a wide variety of learning algorithms. For instance, in the binary classi cation setting it includes the Perceptron algorithm [28, 4, 27] and algorithms in the Winnow family [23, 24] such as Weighted Majority; in the regression setting it includes the Widrow Ho rule [33] and algorithms in the EG family [21] All these algorithms have the same basic structure. In the generic trial t the algorithm stores the weight vector w t , ....

Block, H. D. (1962), The perceptron: A model for brain functioning, Reviews of Modern Physics, 34, 123-135. Reprinted in Neurocomputing by Anderson and Rosenfeld.

....is xed at 0. Keywords: Online Learning, Large Margin Classi ers, Perceptrons, Support Vector Machines. 1. Introduction The perceptron algorithm [33, 34] and the maximum margin classi er [3] have similar theoretical bases, but di erent strengths. In the case of linearly separable data, Block [2], Noviko [29] and Minsky and Papert [27] showed that the number of mistakes made by the perceptron algorithm is upper bounded by a function of the margin, i.e. the minimal distance from any instance to the separating hyperplane. Freund and Schapire [9] generalized this result to the inseparable ....

Block, H. D.: 1962, `The perceptron: A model for brain functioning'. Reviews of Modern Physics 34, 123-135.

....1 ) x m ; y m ) of patternclassi cation pairs in R n f1; 1g, let R = max t jj x t jj. If there is a weight vector u such that jj ujj = 1 and y t ( u x t ) for all 1 t m, then the number of mistakes made by the perceptron algorithm on this sequence is at most R 2 = 2 [2, 23]. The idea that a large margin classi er might be expected to give better generalization performance can be partially motivated by the above mistake bound. Vapnik s optimal margin classi er [3] which classi es all the training data correctly and maximizes the margin, i.e. the minimal distance ....

Block, H. D.: 1962, `The perceptron: A model for brain functioning'. Reviews

....process is easy to understand as the result of a simple mathematical formula. We discuss this property in more detail in Section 5.7. 3.2 How Perceptrons Work The perceptron was introduced in 1962 [19] as a way to study brain function. We consider the simplest of many types of perceptrons [2], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function t(x1 ; xn) of n inputs. In our case, the x i are the bits of a global branch history shift register, and the ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

....and by NSF grant CCR 95 04436. 1 Introduction Learning an unknown linear threshold function from labeled examples is one of the most fundamental problems in computational learning theory. Many algorithms have been analyzed in the noise free setting where all examples are labeled correctly (see [8, 10, 25] for some representative approaches) and powerful algorithms have recently been developed for the classification noise model where the labels of examples may be corrupted [9, 12] While such algorithms and analyses are theoretically interesting, their utility for practical learning problems is ....

....threshold ; i.e. f( x) sign( u Delta x Gamma ) but our definition incurs no real loss of generality since such a threshold can be simulated by adding an extra variable. 2. 1 Perceptron The Perceptron algorithm for learning linear threshold functions was introduced nearly forty years ago [8, 27, 28] and continues to be the subject of active theoretical research [7, 9, 16, 23, 30, 31] The algorithm works online and is remarkably simple: it maintains a prediction vector v 2 n which is initially set to zero. Given an example x 2 n ; the algorithm predicts y = sign( v Delta x) 2 ....

[Article contains additional citation context not shown here]

H. Block. The perceptron: a model for brain functioning, Reviews of Modern Physics 34 (1962), 123-135.

....a branch instruction is encountered, the address of the branch is hashed to select a neuron from the table. The selected neuron is used to predict the likely direction of the branch. Once the actual branch outcome is known, the neuron is updated with the training rule. We use the Block neuron [11], a type of perceptron [12, 13] because it provides good accuracy and lends itself to a fast and compact representation in hardware. Although perceptrons cannot learn linearly inseparable functions with 100 accuracy [14] we have found that they work well in practice for branch prediction. The ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123-135, 1962.

....process is easy to understand as the result of a simple mathematical formula. We discuss this property in more detail in Section 5.6. 3.2 How Perceptrons Work The perceptron was introduced in 1962 [20] as a way to study brain function. We consider the simplest of many types of perceptrons [3], a single layer perceptron consisting of one artificial neuron connecting several input units by weighted edges to one output unit. A perceptron learns a target Boolean function t(x 1 ; xn ) of n inputs. In our case, the x i are the bits of a global branch history shift register, and the ....

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

No context found.

H. D. Block. The Perceptron: A model for brain functioning. Reviews of Modern Physics, 34: 123--135, 1962.

No context found.

H.D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962. Reprinted in Neurocomputing, Anderson and Rosenfeld.

No context found.

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

No context found.

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

No context found.

BLOCK, H. D. 1962. The perceptron: A model for brain functioning. Rev. Mod. Phy. 34, 123--135.

No context found.

H. D. Block. The perceptron: A model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

No context found.

H.D. Block. The Perceptron: a model for brain functioning. Review of Modern Physics, 34:123135, 1962.

No context found.

H. Block. The perceptron: a model for brain functioning. Reviews of Modern Physics, 34:123{ 135, 1962.

No context found.

H. Block. The Perceptron: a model for brain functioning. Reviews of Modern Physics, 34:123--135, 1962.

First 50 documents

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