Ex Parte Ensel et al

Board of Patent Appeals and Interferences

Sep 10, 2010

10042278 (B.P.A.I. Sep. 10, 2010)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
10/042,278 01/11/2002 Christian Ensel 1454.1212 5517
21171 7590 09/10/2010
STAAS & HALSEY LLP
SUITE 700
1201 NEW YORK AVENUE, N.W.
WASHINGTON, DC 20005
EXAMINER
CHANKONG, DOHM
ART UNIT PAPER NUMBER
2452
MAIL DATE DELIVERY MODE
09/10/2010 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
____________________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
____________________

Ex parte CHRISTIAN ENSEL and VOLKMAR STERZING
____________________

Appeal 2009-006326
Application 10/042,2781
Technology Center 2400
____________________

Before JEAN R. HOMERE, DEBRA K. STEPHENS, and
JAMES R. HUGHES, Administrative Patent Judges.

HUGHES, Administrative Patent Judge.

DECISION ON APPEAL2

1 Application filed January 11, 2002. The real party in interest is Siemens
Aktiengesellschaft. (App. Br. 1.)
2 The two-month time period for filing an appeal or commencing a civil
action, as recited in 37 C.F.R. § 1.304, or for filing a request for rehearing,
as recited in 37 C.F.R. § 41.52, begins to run from the “MAIL DATE”
(paper delivery mode) or the “NOTIFICATION DATE” (electronic delivery
mode) shown on the PTOL-90A cover letter attached to this decision.
Appeal 2009-006326
Application 10/042,278

2
STATEMENT OF THE CASE
Appellants appeal from the Examiner’s rejection of claims 1-11, 13,
14, and 27-29 under authority of 35 U.S.C. § 134(a). Claims 12 and 15-26
have been canceled. The Board of Patent Appeals and Interferences (BPAI)
has jurisdiction under 35 U.S.C. § 6(b).
We affirm.
Appellants’ Invention
Appellants invented a device, medium, and method for monitoring of
a telecommunication network and training of a statistical estimator for
monitoring of a telecommunication network. (Spec. 1, ¶ [0002].)3
Representative Claim
Independent claims 1 further illustrates the invention. It reads as
follows:
1. A method for computer-aided monitoring of a
telecommunication network formed of devices capable of
communication, said method comprising:
determining training activity parameters, each describing
activity of at least one of a corresponding device and a
corresponding service;
determining possible dependences between devices and
services from the training activity parameters;
determining from the possible dependences a normal
range of dependence for at least some of the devices and
services in essentially undisturbed states to train a neural
network as a statistical estimator;

3 We refer to Appellants’ Specification (“Spec.”); Appeal Brief (“App. Br.”)
filed January 29, 2008; and Reply Brief (“Reply Br.”) filed June 23, 2008.
We also refer to the Examiner’s Answer (“Ans.”) mailed April 21, 2008.
Appeal 2009-006326
Application 10/042,278

3
determining current activity parameters, each describing
activity of at least one of a corresponding device and a
corresponding service;
comparing the current activity parameters by the
statistical estimator with the normal range of dependence; and
determining from said comparing whether at least one of
the devices and services in the telecommunication network has
a communication performance different from the normal range
of dependence in accordance with a predetermined criterion.

References
The Examiner relies on the following references as evidence of
unpatentability:
Waclawsky US 5,974,457 Oct. 26, 1999

Nittida Nuansri, Tharam S. Dillon & Samar Singh, An Application of
Neural Network and Rule-Based System for Network Management:
Application Level Problems, Proceedings of the Thirtieth Hawaii
International Conference on System Sciences (HICSS) Vol. 5, 474-483
(January 1997) (hereinafter “Nuansri”).

Rejection on Appeal
The Examiner rejects claims 1-11, 13, 14, and 27-29 under 35 U.S.C.
§ 103(a) as being unpatentable over the combination of Waclawsky and
Nuansri.

ISSUE
Based on our review of the administrative record, Appellants’
contentions, and the Examiner’s findings and conclusions, the pivotal issue
before us is as follows:
Appeal 2009-006326
Application 10/042,278

4
Does the Examiner err in finding the Waclawsky and Nuansri
references can properly be combined, and would have collectively taught or
suggested “determining from the possible dependences a normal range of
dependence for at least some of the devices and services in essentially
undisturbed states to train a neural network as a statistical estimator” as
recited in Appellants’ claim 1?

FINDINGS OF FACT (FF)
Appellants’ Specification
1. Appellants’ Specification describes a method for monitoring a
telecommunication network, including: determining activity parameters
describing activity between a device and a corresponding service, e.g.,
communication activity among devices or services over a communication
network; determining possible dependences (relationships and/or network
connections) between devices and services from the activity parameters; and
determining from the possible dependences a normal range of dependence
(normal/usual performance) for at least some of the devices and services in
essentially undisturbed states (in normal operation (i.e., not under attack –
see ¶¶ [0005], [0008]-[0010], [0018]-[0022])) to train a neural network as a
statistical estimator. (¶¶ [0005]-[0010], [0018]-[0022], [0024], [0025],
[0038]-[0041], [0045]-[0051], & [0077].)
2. The described devices may include computers (personal
computers, workstations, laptops, and servers), printers and/or switches.
(¶ [0006].) The described services may include “application programs in a
state of execution[,] . . . for example, a web server, a file server, databases,
[and] various application servers.” (¶ [0007].) The Specification explains
Appeal 2009-006326
Application 10/042,278

5
that activity and activity parameters (training activity parameters) may
include: the number of data packets transmitted or received by a device or
service, “the processor utilization of the respective device,” and/or “the
number of predetermined system function calls” by the device or service.
(¶ [0050]; see ¶¶ [0020], [0039].) The Specification also explains that
dependences are “communication-related dependences between the devices
or services” (¶ [0040]) or “combinations of two devices or services coupled
to one another in the telecommunication network” (¶ [0045]). (See
¶¶ [0038]-[0041], & [0045]-[0049].) The “normal range of dependence” is
equivalent to the “normal performance” or “undisturbed performance” of the
devices/services, i.e., a normal “communication performance” that is within
“a predetermined range of tolerances.” (¶¶ [0018], [0021], [0022].) The
Specification further explains that “[t]he statistical estimator is trained with
the usual performance of the devices or services, that is to say with the
normal range of dependence.” (¶ [0025]; see [0024]-[0025], [0038]-[0041],
& [0045]-[0049].) In an embodiment of Appellants’ invention, the system
stores the devices, services, and activity parameters in a database (matrix),
and determines from the matrix the device/service dependences (pairs) and
normal range of dependence (operation). (Spec. ¶¶ [0047]-[0049].)
3. Appellants do not explicitly define a “neural network” in their
Specification. The Specification, however, does describe the functionality
of the neural network – the neural network “makes it possible to model both
local relationships and global relationships of the communication
performance of the respective pair of devices” (¶ [0077]) – as well as an
exemplary neural network. (¶¶ [0071]-[0077]; Fig. 2.) Appellants define a
“statistical estimator” as “a basically arbitrary neural model, that is to say a
Appeal 2009-006326
Application 10/042,278

6
neural network, or else a neuro-fuzzy model, which is trained by known
training methods.” (¶ [0051].)
Waclawsky Reference
4. Waclawsky describes:
[A] system and method to enable real-time establishment and
maintenance of a standard of operation for a data
communications network. The standard is a data set which
includes network activity which is historically categorized by
traffic type and by activity. The process begins with monitoring
the network media or some network component over some
period of time. The monitoring information is used to build
benchmark data sets. The benchmark data sets contain a
standard of operation for the network, which are historically
categorized by either traffic type or activity. This standard of
operation is constantly built by the intelligent monitoring
facilities. After some period of time which is referred to as the
benchmark data set refresh interval, the benchmark that was
created is employed in a fashion to allow a determination as to
whether the data that is taken from the current monitoring
activity indicates normal network behavior.
(Abstract.) Waclawsky further describes the operation of its expert system
(the expert analysis module) and intelligent monitoring system (the
intelligent realtime monitor) – the expert system compares communication
network data inputs (event vectors) to network behavior standards (rules) for
determining if the communication network data falls within normal
parameters, and the intelligent monitoring system produces benchmark data
sets that may be substituted for the standards used by the expert system:
[T]he expert analysis module 160, will have the information
represented by the event vectors compared with standards of
behavior for the network. The comparison is performed by the
rules contained in the rule based criteria modules 150. The
standards can be predetermined, predefined standards such as
Appeal 2009-006326
Application 10/042,278

7
average utilization for particular types of traffic such as batch
traffic, interactive traffic, voice traffic or video traffic. Another
important type of standard is the benchmark data set which is
the accumulated history of behavior of traffic on the network,
as it has been monitored by the system shown in FIGS. 1A and
1B. The benchmark data sets 110, in accordance with the
invention, can progressively accumulate a more accurate
representation of the expected behavior for the network and that
standard can be substituted for the predetermined standard used
by the rules in the rule based criteria modules 150.
(Col. 4, l. 56 to col. 5, l. 4; see Figs. 1A-1B.)
5. Waclawsky explains that the expert system compares the
communication network data against predetermined standards and/or
benchmark data sets to produce outputs (inference signals) that represent the
behavior of the network or a particular user defined class of communication
traffic – such as particular nodes, packet sizes, and/or types of traffic (batch,
video, or audio/voice traffic) – and may be used to control the network.
(Col. 5, ll. 5-30.) Waclawsky also describes that the benchmark data set
may describe a range of ordinary network behavior that falls within certain
limits. (Col. 10, l. 54 to col. 11, l. 23.) For example, Waclawsky describes
determining and comparing benchmark triggers (peak utilization range and
frame/second range) (See Fig. 10B).
Nuansri Reference
6. Nuansri describes a hybrid network management system that
includes a neural network and a rule-based system for monitoring and
diagnosing problems in a communication network, in particular, at an
application level. (p. 474, Abstract & Sec. 1.) Nuansri’s neural network
(BRAINNE) has the capability to “learn complex, non-linear functions.”
(p. 475, para. 1; see p. 477, Sec. 5.) The rule-based expert system monitors,
Appeal 2009-006326
Application 10/042,278

8
for example, the domain name system (DNS) – which utilizes name servers
and a distributed database to provide DNS services (resolve names and IP
addresses) (p. 475) – and the neural network (BRAINNE) acquires
knowledge concerning, for example, application (DNS) errors and the causes
of these errors. (pp. 475, Sec. 2; 477-78, Sec. 5.) The neural network
(BRAINNE) creates rules and provides these to the expert system, and the
rule-based expert system analyzes and diagnoses problems in the
communication system. (p. 478, para. 1.)

ANALYSIS
Appellants contend that the Waclawsky and Nuansri references “taken
alone or in combination, fail to describe ‘determining from the possible
dependences a normal range of dependence for at least some of the devices
and services essentially undisturbed states to train a neural network as a
statistical estimator’ as recited by claim 1” (Reply Br. 1). (See App. Br. 8-
10; Reply Br. 1-3.) Appellants also contend that the Examiner’s cited
combination of the Waclawsky and Nuansri references is improper “because
there is insufficient evidence for a motivation to use the method of Nuansri
et al. in the system described by Waclawsky et al.” (App. Br. 11), and that
Waclawsky “teaches away from a statistical estimator training system”
(App. Br. 9). (See App. Br. 9, 10-11; Reply Br. 3.) The Examiner finds that
the Waclawsky and Nuansri references teach each feature of Appellants’
claim 1 and maintains that the claim is properly rejected as obvious over the
reference combination. (Ans. 3-5, 8-12.) Specifically, the Examiner finds
that: Waclawsky describes “determining a normal range of dependence”
(Ans. 9), and “training a rule-based network as a statistical estimator” (Ans.
Appeal 2009-006326
Application 10/042,278

9
10); Nuansri describes a hybrid neural network and rule-based network
combination, as well as training the neural network (Ans. 10); and that one
skilled in the art at the time of Appellants’ invention would have been
motivated to make the cited reference combination (Ans. 11-12). Based on
these contentions, we decide the question of whether the Examiner erred in
finding the Waclawsky and Nuansri references can properly be combined,
and would have collectively taught or suggested the disputed feature of
determining from the possible dependences a normal range of dependence
for at least some of the devices and services in essentially undisturbed states
to train a neural network as a statistical estimator.
After reviewing the record on appeal, we agree with the Examiner that
a skilled artisan could have properly combined the Waclawsky and Nuansri
references and that the combination would have taught the disputed feature.
The dispute before us hinges on the disagreement of the Examiner and
Appellants as to what constitutes “possible dependences” (between devices
and services), “a normal range of dependence,” “a neural network,” and “a
statistical estimator,” and the interpretation of these terms is critical to
resolving this dispute. Thus, we begin our analysis by construing
Appellants’ disputed claim limitation. We give claim terminology the
“broadest reasonable interpretation consistent with the [S]pecification” in
accordance with our mandate that “claim language should be read in light of
the [S]pecification as it would be interpreted by one of ordinary skill in the
art.” In re Am. Acad. of Sci. Tech Ctr., 367 F.3d 1359, 1364 (Fed. Cir. 2004)
(citations omitted).
Appeal 2009-006326
Application 10/042,278

10
Appellants’ claim 1 in relevant part recites:
determining training activity parameters, each describing
activity of at least one of a corresponding device and a
corresponding service; determining possible dependences
between devices and services from the training activity
parameters; [and] determining from the possible dependences a
normal range of dependence for at least some of the devices
and services in essentially undisturbed states to train a neural
network as a statistical estimator.
(App. Br. 12, Claim App’x., claim 1 (emphasis added to the limitation in
dispute).) Appellants describe “dependences” as “communication-related”
relationships between computer devices (or peripherals), and/or services
such as databases and application programs executed by servers.
Appellants’ system determines the dependences from communication
network activity (activity parameters) including traffic volume (data packets
transmitted/received), processor utilization, and system function calls.
Appellants explicitly describe the “normal range of dependence” as
equivalent to the normal or usual communication performance of a particular
device/service or device/service pair. (FF 1-2.) Appellants explicitly define
a “statistical estimator” as “a basically arbitrary neural model, that is to say a
neural network, or else a neuro-fuzzy model, which is trained by known
training methods.” Appellants do not explicitly define a “neural network”
either in their claim or Specification. (FF 3.) We understand a neural
network to be a system of programs and data structures that approximates
the operation of a biological neural network (the human brain). A neural
network typically “learns” (is trained) by receiving weighted inputs (data
and rules about data relationships). A neural network adapts with time and
repetition to produce appropriate outputs, i.e., it “learns” from examples.
See Microsoft Computer Dictionary, Fifth Edition (May 2002); Free On-line
Appeal 2009-006326
Application 10/042,278

11
Dictionary of Computing (“FOLDOC”), available at http://foldoc.org. We
also understand a “statistical estimator” to be a device or program, in this
instance a neural network, that performs the function of providing an
estimate (approximate calculation) based on statistical analysis.
Although ultimately unnecessary for our construction of the disputed
claim limitation, we also note it is known within the art (as argued at length
by Appellants – see App. Br. 7) that an expert system uses sets of rules and
data to produce a decision or recommendation. Neural networks, on the
other hand, attempt to simulate the human brain by collecting and processing
data for the purpose of “learning,” and adapt their analyses criteria through
the learning process. One of the generally recognized differences between
an expert system and a neural network is that a neural network can adapt its
criteria to better match the data it analyzes, while an expert system typically
produces results without adjusting for changes in the analyzed data. See
generally, Microsoft Computer Dictionary; FOLDOC; App. Br. 7-8.
Based on Appellants’ disclosures, we broadly but reasonably construe
Appellants’ disputed claim limitation to mean determining, based on
communication network activity, a normal communication performance or
range of performance, and providing the normal performance as an input to a
neural network – a system of programs and data structures that learns and
adapts with time and repetition – to train the neural network to provide an
estimate based on statistical analysis (function as a statistical estimator).
This construction is consistent with the cited references, Appellants’
Specification, and the knowledge of those skilled in the art at the time of
Appellants’ invention.
Appeal 2009-006326
Application 10/042,278

12
As detailed in the Findings of Fact section supra, the Waclawsky
reference describes a system for determining a standard of operation for a
data communications network, portions of the network, and/or particular
nodes, types of communication traffic, or activities in the network.
Waclawsky’s system monitors portions of the network, such as particular
communications links or components to build benchmark data sets that the
system utilizes as the standard of operation, in addition to predefined rules,
to determine if the current operation of the network is behaving normally –
that is, within a normal range of network behavior. (FF 4-5.) Waclawsky’s
benchmark data sets also describe a range of ordinary network behavior that
falls within certain limits, for example, benchmark triggers including peak
utilization range and frame/second range. (FF 5.) Accordingly, Waclawsky
describes monitoring communication network activity between network
devices to produce benchmark data, and an adaptive expert system that
adapts its evaluation criteria to include benchmark data sets for a range of
ordinary activity. This range of normal activity is equivalent to an estimate
of normal activity produced by statistical analysis. Thus, we find
Waclawsky teaches determining, based on communication network activity,
a normal communication performance or range of performance, and
providing the normal performance as an input to an adaptive expert system
(a system of programs and data structures that learns and adapts with time
and repetition) which functions as a statistical estimator. We also find
Waclawsky teaches training its adaptive expert system utilizing the
benchmark data.
The Nuansri reference describes a hybrid communication network
management system including a neural network and a rule-based system for
Appeal 2009-006326
Application 10/042,278

13
monitoring and diagnosing problems in a communication network, in
particular, at an application level. Nuansri’s neural network (BRAINNE)
and rule-based expert system monitors communication devices and services,
for example, the domain name system (DNS) which utilizes servers and a
distributed database to provide DNS services. The neural network creates
evaluation rules and/or criteria, and provides these to the rule-based expert
system, which analyzes and diagnoses problems in the communication
system. (FF 6.) Accordingly, we find Nuansri teaches monitoring
communication activity between network devices and/or services, a hybrid
system including a neural network and an expert system, and training a
neural network utilizing monitored communication network activity.
We find Appellants’ contrary arguments unpersuasive. Specifically,
Appellants mischaracterize Waclawsky, Nuansri, and the combination of
Waclawsky and Nuansri. Specifically, Appellants mischaracterize
Waclawsky as merely describing a rule-based system that teaches away from
a neural network (App. Br. 9; Reply Br. 2, 3), and further state that
Waclawsky does not “describe how the ‘benchmark data set’ could be used
as a ‘normal range of dependence’” (App. Br. 10). Appellants
mischaracterize Nuansri’s neural network as merely a “pre-processor.”
(App. Br. 8, 9; Reply Br. 2.) We find (supra) that Waclawsky teaches a
normal range of network performance, and Nuansri teaches training a neural
network utilizing network communication data.
Thus, we find the cited reference combination of Waclawsky and
Nuansri teaches Appellants’ disputed claim limitation as recited in
Appellants’ independent claim 1. Appellants do not separately argue
dependent claims 2-11, 13, and 14 (dependent on claim 1), or independent
Appeal 2009-006326
Application 10/042,278

14
claims 27, 28, and 29. Accordingly, we select independent claim 1 as
representative of claims 2-11, 13, 14, and 27-29, and we find the Waclawsky
and Nuansri references render these claims obvious for the reasons set forth
with respect to representative claim 1. It follows that Appellants do not
persuade us of error in the Examiner’s obviousness rejection of claims 1-11,
13, 14, and 27-29, and we affirm the Examiner’s rejection of these claims.

CONCLUSIONS OF LAW
Appellants have not shown that the Examiner erred in rejecting claims
1-11, 13, 14, and 27-29 under 35 U.S.C. § 103(a).

DECISION
We affirm the Examiner’s rejection of claims 1-11, 13, 14, and 27-29
under 35 U.S.C. § 103(a).

No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED

msc

Staas & Halsey LLP
Suite 700
1201 New York Avenue, N.W.
Washington, DC 20005