161.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the available resources of (a) abundant computing power and (b) domain-specific knowledge to improve its ability to generalize. Connectionist theory-refinement systems, which use background knowledge to select a neural network's topology and initial weights, have proven to be effective at exploiting domain-specific knowledge; however, most do not exploit available computing power. This weakness occurs because they lack the ability to refine the topology of the neural networks they produce, thereby limiting generalization, especially when given impoverished domain theories. We present the Regent algorithm which uses...
162.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
163.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
164.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
165.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
166.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
167.
Connectionist Theory Refinement: Genetically Searching the Space of Network Topologies
- David W. Opitz,Jude W. Shavlik
An algorithm that learns from a set of examples should ideally be able to exploit the
available resources of (a) abundant computing power and (b) domain-specific knowledge to
improve its ability to generalize. Connectionist theory-refinement systems, which use background
knowledge to select a neural network's topology and initial weights, have proven to
be effective at exploiting domain-specific knowledge; however, most do not exploit available
computing power. This weakness occurs because they lack the ability to refine the
topology of the neural networks they produce, thereby limiting generalization, especially
when given impoverished domain theories. We present the Regent algorithm which uses
(a) domain-specific knowledge to help create an initial...
168.
Virtual Topologies: A New Concurrency Abstraction for High-Level Parallel Languages
- James Philbin,Suresh Jagannathan,Rajiv Mirani
ion for
High-Level Parallel Languages
(Preliminary Report)
James Philbin
1
, Suresh Jagannathan
1
, Rajiv Mirani
2
1
Computer Science Division, NEC Research Institute, 4 Independence
Way, fphilbin|sureshg@research.nj.nec.com
2
Department of Computer Science, Yale University, New Haven, CT
mirani@cs.yale.edu
Abstract. We present a new concurrency abstraction and implementation technique
for high-level (symbolic) parallel languages that allows significant programmer
control over load-balancing and mapping of fine-grained lightweight threads.
Central to our proposal is the notion of a virtual topology. A virtual topology
defines a relation over a collection of virtual processors, and a mapping of those
processors to a set of physical processors; processor topologies configured as
trees, graphs, butterflies, and meshes are some well-known examples. A virtual
topology need not have...
169.
Polishness of Weak Topologies Generated by Gap and Excess Functionals
- Roberto Lucchetti
Introduction
Let (X; d) be a separable complete metric space and CL(X) the hyperspace of X, i.e.
the space of all nonempty closed subsets of X. We are interested to find a class of Polish
topologies on CL(X). In [2] it is shown that the Wijsman topology generated by the
metric d is Polish and in [6] is proved that also the Wijsman topology generated by a
metric % topologically equivalent to d is Polish.
We are going to show that, if Delta is a subfamily of CLB(X) containing the singletons
and separable with respect to the induced Hausdorff metric , then the weak topology
G
Delta
generated by the...
170.
Structured Topological Complexes: A Feature-based API for Non-Manifold Topologies
- Jarek Rossignac
Much of recent research in representation schemes for solid
modeling was focused on the extension of boundary representations
to support non-manifold topologies. We introduce here a
very simple, yet general formalism, which subsumes and simplifies
many previous attempts at defining high-level operators
for creating and interrogating such models. Our Structured
Topological Complex (STC) extends our previous work on
Selective Geometric Complexes (SGC) and on Constructive
Non-Regularized Geometry (CNRG) and provides the foundations
for a new generation of representations-both constructive
and evaluated-and of APIs that are independent of the particular
geometric domain and even of the particular approximation
scheme for geometric primitives.
1. INTRODUCTION
In spite of recent progress in non-manifold topological modeling
technologies (see the...
171.
Quasinormable Spaces And The Problem Of Topologies Of Grothendieck
- Scientiarum Fennicae,Alfredo Peris
. This article is dedicated to the study of quasinormable injective tensor products
of locally convex spaces and quasinormable spaces of continuous linear operators. The stability of
the quasinormability is obtained in the frame of the class of spaces which are quasinormable by
operators; this class, introduced and studied here, contains many function spaces. The problems
considered in the article are closely related to the problem of topologies of Grothendieck. A
characterization of the quasinormable spaces which are (FBa)-spaces in the sense of Taskinen
is obtained and new examples and counterexamples are given. In particular we show that the
quasinormable space l p+ is a concrete example...
172.
Design of Logical Topologies for Wavelength-Routed Optical Networks
- Rajiv Ramaswami,Kumar N. Sivarajan
This paper studies the problem of designing a logical topology over a wavelengthrouted
all-optical network physical topology. The physical topology consists of the nodes
and fiber links in the network. On an all-optical network physical topology, we can set
up lightpaths between pairs of nodes, where a lightpath represents a direct optical
connection without any intermediate electronics. The set of lightpaths along with the
nodes constitutes the logical topology. For a given network physical topology and traffic
pattern (relative traffic distribution among the source-destination pairs), our objective
is to design the logical topology and the routing algorithm on that topology so as to
minimize the network congestion while...
173.
Pattern Matching and Pattern Discovery Algorithms for Protein Topologies
- David Gilbert
. We describe algorithms for pattern matching and pattern
learning in TOPS diagrams (formal descriptions of protein topologies).
These problems can be reduced to checking for subgraph isomorphism
and finding maximal common subgraphs in a restricted class of ordered
graphs. We have developed a subgraph isomorphism algorithm for
ordered graphs, which performs well on the given set of data. The
maximal common subgraph problem then is solved by repeated
subgraph extension and checking for isomorphisms. Despite the
apparent inefficiency such approach gives an algorithm with time
complexity proportional to the number of graphs in the input set and is
still practical on the given set of data. As a result...
174.
Efficient Reinforcement Learning through Evolving Neural Network Topologies
- Kenneth O. Stanley,Risto Miikkulainen
Neuroevolution is currently the strongest method on the pole-balancing benchmark reinforcement learning tasks. Although earlier studies suggested that there was an advantage in evolving the network topology as well as connection weights, the leading neuroevolution systems evolve fixed networks. Whether evolving structure can improve performance is an open question. In this article, we introduce such a system, NeuroEvolution of Augmenting Topologies (NEAT). We show that when structure is evolved (1) with a principled method of crossover, (2) by protecting structural innovation, and (3) through incremental growth from minimal structure, learning is significantly faster and stronger than with the best fixed-topology...
175.
Inference and Labeling of Metric-Induced Network Topologies
- Azer Bestavros,John Byers,Khaled Harfoush
The development and deployment of distributed networkaware
applications and services require the ability to compile and maintain
a model of the underlying network resources with respect to (one or more)
characteristic properties of interest. To be manageable, such models must be
compact, and to be general-purpose, they should enable a representation of
properties along temporal, spatial, and measurement resolution dimensions.
In this paper, we propose MINT---a general framework for the construction
of such metric-induced models using end-to-end measurements. We present
the basic theoretical underpinnings of MINT for a broad class of metrics
obeying certain properties. We instantiate MINT for two metrics of interest,
namely packet loss rates and bottleneck bandwidth....
176.
Two-Loop Master Integrals for ... Jets: The planar topologies
- T. Gehrmann
The calculation of the two-loop corrections to the three jet production rate and to event shapes in
electron-positron annihilation requires the computation of a number of up to now unknown two-loop
four-point master integrals with one o-shell and three on-shell legs. In this paper, we compute those
master integrals which correspond to planar topologies by solving dierential equations in the external
invariants which are ful
lled by the master integrals. We obtain the master integrals as expansions
in = (4 d)=2, where d is the space-time dimension. The results are expressed in terms of newly
introduced two-dimensional harmonic polylogarithms, whose properties are shortly discussed. For all
two-dimensional harmonic...
177.
Inference and Labeling of Metric-Induced Network Topologies
- Azer Bestavros,John Byers,Khaled Harfoush
The development and deployment of distributed
network-aware applications and services require the ability
to compile and maintain a model of the underlying network
resources with respect to (one or more) characteristic
properties of interest. To be manageable, such models
must be compact, and to be general-purpose, they should
enable a representation of properties along temporal, spatial,
and measurement resolution dimensions. In this paper,
we propose MINT|a general framework for the construction
of such metric-induced models using end-to-end measurements.
We present the basic theoretical underpinnings
of MINT for a broad class of metrics obeying certain properties.
We instantiate MINT for two metrics of interest,
namely packet loss rates and bottleneck bandwidth. For the
loss rate...
178.
Inference and Labeling of Metric-Induced Network Topologies
- Azer Bestavros,John Byers,Khaled Harfoush
The development and deployment of distributed networkaware
applications and services require the ability to compile and maintain
a model of the underlying network resources with respect to (one or more)
characteristic properties of interest. To be manageable, such models must be
compact, and to be general-purpose, they should enable a representation of
properties along temporal, spatial, and measurement resolution dimensions.
In this paper, we propose MINT---a general framework for the construction
of such metric-induced models using end-to-end measurements. We present
the basic theoretical underpinnings of MINT for a broad class of metrics
obeying certain properties. We instantiate MINT for two metrics of interest,
namely packet loss rates and bottleneck bandwidth....
179.
Automatic Indexing, Retrieval and Reuse of Topologies in Architectual Layouts
- C-h. Coulon
Former layouts contain much of the know-how of architects. A generic and automatic way
to formalize this know-how in order to use it by a computer would save a lot of effort and
money. However, there seems to be no such way. The only access to the know-how are the
layouts themselves. Developing a generic software tool to reuse former layouts you cannot
consider every part of the architectual domain or things like personal style. Tools used today
only consider small parts of the architectual domain. Any personal style is ignored. Isn't it
possible to build a basic tool which is adjusted by the content of...
180.
Penguin Topologies, Rescattering Effects and Penguin Hunting with
- Andrzej J. Buras,Robert Fleischer,Thomas Mannel
In the recent literature, constraints on the CKM angle fl arising from the branching
ratios for B
Sigma
! ß
Sigma
K and B d ! ß
Upsilon
K
Sigma
decays received a lot of attention. An
important theoretical limitation of the accuracy of these bounds is due to rescattering
effects, such as B
+
! fß
0
K
+
g ! ß
+
K
0
. We point out that these processes
are related to penguin topologies with internal up quark exchanges and derive
SU(2) isospin relations among the B
+
! ß
+
K
0
and B
0
d
! ß
Gamma
K
+
decay amplitudes
by defining "tree" and "penguin" amplitudes in a proper way, allowing the
derivation of generalized bounds on the CKM angle fl. We propose strategies to
obtain insights into the dynamics of...
