Ex Parte Rundle et al

Patent Trial and Appeal Board

Dec 15, 2014

11438720 (P.T.A.B. Dec. 15, 2014)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte ALFRED T. RUNDLE and LENNART A. SAAF
____________

Appeal 2012-003629
Application 11/438,720
Technology Center 2600
____________

Before JAMES T. MOORE, JOHN A. JEFFERY, and
DENISE M. POTHIER, Administrative Patent Judges.

POTHIER, Administrative Patent Judge.

DECISION ON APPEAL
STATEMENT OF THE CASE
Appellants appeal under 35 U.S.C. § 134(a) from the Examiner’s
rejection of claims 1–15.1 We have jurisdiction under 35 U.S.C. § 6(b). We
affirm.

1 Throughout this opinion, we refer to the Appeal Brief (“App. Br.”) filed
October 11, 2011; (2) the Examiner’s Answer (“Ans.”) mailed November 2,
2011; and (3) the Reply Brief (“Reply Br.”) filed December 30, 2011.
Appeal 2012-003629
Application 11/438,720

2
Invention
Appellants’ invention relates to a method and system for arbitrating
competing results from systems, such as address recognition systems. Data
characteristics are derived from the results, and the results and
characteristics are used in rules, which are in turn used to arbitrate between
the results. See Spec. ¶¶ 5–6, 14. Illustrative claim 1 is reproduced below
with emphasis:
1. A computer implemented method to arbitrate between results to be
arbitrated in order to effectuate a desired result, each of said results to
be arbitrated obtained by one of a plurality of procedures from
observationally derived data, each of said results to be arbitrated
having a representative characteristic, said representative
characteristic hereinafter referred to as a type, said method comprising
the steps of:
a. ranking, utilizing a processor and computer usable memory that
causes the processor to perform the method, each of the procedures
according to confidence on each of said procedures;
b. classifying each of the results to be arbitrated, utilizing a
processor and computer usable memory that causes the processor to
perform the method, according to a level of accuracy at which each of
said results to be arbitrated matches a datum from a database of a
plurality of databases;
c. comparing, utilizing a processor and computer usable memory
that causes the processor to perform the method, each of the results to
be arbitrated to each other;
d. obtaining, utilizing a processor and computer usable memory
that causes the processor to perform the method, at least one indicator
for each result to be arbitrated, said at least one indicator indicating a
degree of success of the step of comparing each of the results to be
arbitrated to each other;
e. generating rules, utilizing a processor and computer usable
memory that causes the processor to perform the method, said rules
utilizing as inputs said results to be arbitrated and information
derived from said results to be arbitrated;
f. selecting at least one of said rules;
Appeal 2012-003629
Application 11/438,720

3
g. applying, utilizing a processor and computer usable memory that
causes the processor to perform the method, said at least one of said
rules to the results to be arbitrated and to the at least one indicator for
each result to be arbitrated, said at least one of said rules being applied
according to an algorithm;
h. selecting, utilizing a processor and computer usable memory
that causes the processor to perform the method, from the application
of said at least one of said rules, said desired result, said desired result
being one of the results to be arbitrated; said results to be arbitrated
being of a same type of result as other said results to be arbitrated.

The Rejections
Claims 1–15 are rejected under 35 U.S.C. § 112, first paragraph, as
failing to comply with the enablement requirement. Ans. 4–17.

THE CONTENTIONS
Concerning claim 1, the Examiner states that step (g) does not comply
with the enablement requirement and also addresses the “generating rules”
step. Ans. 5–6. Specifically, the Examiner finds that the disclosure
describes generically three types of artificial intelligence used to generate
rules, but the disclosure provides insufficient details of how these described
techniques accomplish the recited “generating rules” step. See Ans. 6–11.
The Examiner further evaluates the claims using the Wands2 factors, and
concludes that the disclosure fails to inform an ordinarily skilled artisan how
to make and use the claimed steps without undue experimentation, when
considering various factors. Ans. 11–17.
Appellants take the opposite position. App. Br. 5–11.

2 In re Wands, 858 F.2d 731 (Fed. Cir. 1988).
Appeal 2012-003629
Application 11/438,720

4
PRINCIPLES OF LAW
“[T]he specification of a patent must teach those skilled in the art how
to make and use the full scope of the claimed invention without ‘undue
experimentation.’” In re Wright, 999 F.2d 1557, 1561 (Fed. Cir. 1993); see
also Wands, 858 F.2d at 737. In reaching the conclusion that
experimentation is undue, many factors must be weighed, including the
quantity of experimentation necessary and the relative skill of those in the
art. Id.

ISSUES
Under § 112, first paragraph, has the Examiner erred in rejecting
claim 1 by finding that the disclosure fails to teach an ordinarily skilled
artisan how to make and use the claimed step of “generating rules, utilizing a
processor and computer usable memory that causes the processor to perform
the method, said rules utilizing as inputs said results to be arbitrated and
information derived from said results to be arbitrated” without undue
experimentation?

ANALYSIS
We select claim 1 as representative. See 37 C.F.R. § 41.37(c)(1)(vii).
Based on the record before us, we find no error in the Examiner’s rejection
of independent claim 1. Although explicitly stating step (g) of independent
claim 1 is the focus of the presented rejection (Ans. 5), the Examiner
addresses mainly the “generating rules” limitation of step (e). Appellants
also indicate step (e) is the step the Examiner concludes does not satisfy the
enablement requirement. Reply Br. 4. Our discussion thus focuses on the
Appeal 2012-003629
Application 11/438,720

5
step of “generating rules, . . . said rules utilizing as inputs said results to be
arbitrated and information derived from said results to be arbitrated” in
claim 1.
As noted by the Examiner, the disclosure states that rules can be
generated by a neural network or genetic algorithm. Ans. 8–11 (citing Spec.
¶ 33). We agree. The Examiner further states the rules can be generated
using an expert system, and Appellants state the rules can be generated using
“other methods.” Ans. 7, 11; App. Br. 5 (both citing Spec. ¶ 33). We do not
necessarily agree that the disclosure supports these additional findings or the
Appellants’ assertion. See Spec. ¶ 33. However, we will focus on the neural
network example in determining whether claim 1 satisfies the enablement
requirement.
In determining whether claim 1 satisfies the enablement requirement,
both the Examiner and Appellants considered at least some of the Wands
factors. Ans. 12–16; App. Br. 8–11. These factors include (1) the quantity
of experimentation necessary, (2) the amount of direction or guidance
presented, (3) the presence or absence of working examples, (4) the nature
of the invention, (5) the state of the prior art, (6) the relative skill of those in
the art, (7) the predictability or unpredictability of the art, and (8) the breadth
of the claims. Wands, 858 F.2d at 737. Accordingly, we apply the Wands
factors in determining whether the disclosure teaches an ordinarily skilled
artisan how to make or use the claimed “generating rules” step.
(1) The Quantity of Experimentation Necessary
The Examiner finds that the amount of experimentation for an
ordinarily skilled artisan to determine how to generate rules using a neural
network or genetic algorithm technique is significant and extensive.
Appeal 2012-003629
Application 11/438,720

6
Ans. 12. Citing to Wlodzislaw Duch et al., Computational Intelligence
Methods for Rule-Based Data Understanding, 92 Procs. of IEEE 771, 801
(2004) (“Duch”), Appellants contend that the amount of experimentation is
relatively simple. App. Br. 9. In the Reply Brief, Appellants restate the
quantity of experimentation as “usual experimentation.” Reply Br. 8.
The Examiner explains that a neural network technique operates on a
function that requires “several inputs, adds them together, performs a
function on the sum, and outputs the result . . . [E]ach input to a neuron has a
separate weighting factor, and the function performed on the sum of the
weighted inputs usually has one or two more parameters which affect the
results.” Ans. 8. The Examiner further refers to Figure 1.1 in the Appeal
Brief reproduced below, showing an exemplary neural network. Ans. 9
(citing App. Br. 6).

Fig. 1.1 of Gallant showing a neural network
The Examiner finds that the parameters are trained by using any of a number
of known search techniques, including back propagation or a genetic
algorithm.3 Ans. 9. Appellants do not dispute these findings.

3 These approaches are further discussed by the Examiner. Ans. 9. See also
Stephen I. Gallant, NEURAL NETWORK LEARNING AND EXPERT SYSTEMS 195-
229 (“Gallant”).
Appeal 2012-003629
Application 11/438,720

7
Appellants state that there are a number of textbooks and publications
on how to generate rules from neural networks. App. Br. 6 (citing Duch and
Gallant at 195-229, 315-304). Appellants add that “once the inputs and
output of the neural network are identified and the training set provided,” an
ordinarily skilled artisan “can generate the neural network and obtain the
rules.” Id. The Examiner does not dispute this assertion. As such, an
ordinarily skilled artisan would need to know, at a minimum, the input(s),
the output(s), and the training set in order to know how to generate rules as
recited in claim 1 using a neural network technique.
For the specific zip code example described in the disclosure (Spec.
¶¶ 8, 21–30), we are provided with inputs. That is, the disclosure states the
rules use as inputs “the results 10 and characteristics derived from the
results, the characteristics including the at least one indicator and the results
of the comparison of results (performed in step 80, Fig. 1).” Spec. ¶ 20;
Fig. 1; see also App. Br. 6 (quoting from this passage). Figure 3 shows the
specific embodiment for generating rules for the zip code embodiment, in
which similar steps 210, 215, 280, and 290 are fed into rule box 20. Spec.,
Fig. 3. Additionally, claim 1 recites explicitly “utilizing as inputs said
results to be arbitrated and information derived from said results to be
arbitrated.” Although the claim does not recite explicitly what “information
derived from the results to be arbitrated” to use as inputs, the disclosure
provides several options, including characteristics of the results (box 215 in

4 Appellants further cite to Wlodzislaw Duch et al., A New Methodology of
Extraction, Optimization and Application of Crisp and Fuzzy Logical Rules,
12 IEEE Trans. Neural Nets. 277 (2001), LiMin Fu, Rule Generation from
Neural Networks, 24 IEEE Trans. On Sys., Man. & Cybernetics 1114
(1994), and others. App. Br. 7.
Appeal 2012-003629
Application 11/438,720

8
Fig. 3), results comparisons from several procedures (box 280 in Fig. 3), and
indicator for degree of success of comparison (box 290 in Fig. 3).5
Accordingly, for the specific zip code embodiment, we find that the
disclosure provides to an ordinarily skilled artisan enough detail to know
what to input into a neural network without undue experimentation.
Concerning what the output should be for the zip code example,
Appellants state that the disclosure indicates the output will be “results most
likely to be accurate . . . .” App. Br. 6 (citing Spec. ¶ 20). Yet, this portion
of the disclosure discusses the outcome of step 110, which is the results of
the algorithm to arbitrate between results corresponding to step 310 in
Figure 3, not the output of the rule generating step at box 20.
For example, the desired results of Rule 1 discussed in paragraph 30
may be to reject any correspondence that contains a “Return to Sender,”
while the overall desired result of algorithm is to have the most accurate
results (i.e., the most accurate zip code). Spec. ¶¶ 20, 30. Also, Rule 1 does
not use the results to be arbitrated (e.g., scanned zip codes) and information
from the results in order to generate a rule that rejects correspondence that
contains “Return to Sender.” This is also true for other rules discussed in
paragraph 30.
Even assuming the disclosure supports that the output discussed in
paragraph 20 is also applicable to the results of the rule generating step, such
that the experimentation to determine the output for the zip code example is
not undue, the disclosure needs to provide a training set to generate the rules.

5 Notably, the results’ characteristics in box 215 is described separately from
boxes 280 and 290, which compare results from other procedures and assign
indicators for degree of success of comparison. Spec., Fig. 3.
Appeal 2012-003629
Application 11/438,720

9
Appellants state that the training set can be obtained according to the
algorithms discussed in paragraph 31. App. Br. 6. Yet, this argument is
circular, given that the algorithms to train the set to obtain rules will use
already-generated rules. See Spec. ¶ 31. As the Examiner indicates, there is
no need to generate rules in this situation using a neural network, because
the rules have already been generated manually. See Ans. 18–19.
Appellants further indicate that an ordinarily skilled artisan can train
the network using techniques, such as back propagation algorithm. App.
Br. 6–7 (citing Gallant at 211–29, 315–30). As such, Gallant may provide a
procedure to develop a training set for the rule generating step that is known
to ordinarily skilled artisan, such that experimentation for the zip code
embodiment is not undue. Appellants cite other publications that cover
similar topics. App. Br. 7–8. Accordingly, we find that the disclosure may
provide enough detail of how to determine the training set for the rule
generating step, such that experimentation is not undue.
However, the disclosure notes “the invention need not be limited to
delivery systems for addressed pieces and could be applied to any system
where arbitration of results between several procedures is required.” Spec.
¶ 8 (italics added). Claim 1 is directed to an invention of arbitrated results
generally. When an ordinarily skilled artisan tries to determine what inputs
(including characteristics of results), outputs, and training sets to use when
generating rules using neural network techniques for a general computer-
implemented method to arbitrate results, the above issues are exacerbated.
The Examiner highlights this issue repeatedly in the Answer. See
Ans. 7 (discussing paragraph 33 having “generic concepts relate[d] to the
rule-generating step of the claim”), 10 (stating the disclosure does not
Appeal 2012-003629
Application 11/438,720

10
address the inputs and outputs for arbitrating among results generically), 11,
17–18 (discussing gaps for adapting the generic concepts, such as neural
networks, to the specific problem at hand), 16 (discussing that paragraph 33
only states rules can be generated using generic concepts, without any
description of how the machine generates the rules).
We note that:
Section 112 requires that the patent specification enable “those
skilled in the art to make and use the full scope of the claimed
invention without 'undue experimentation'” in order to extract
meaningful disclosure of the invention and, by this disclosure,
advance the technical arts. Koito Mfg. [Co., Ltd. v. Turn-Key-
Tech, LLC], 381 F.3d [1142] at 1155 [(Fed. Cir. 2004)]
(quoting Genentech, Inc. v. Novo Nordisk A/S, 108 F.3d 1361,
1365 (Fed. Cir. 1997) (citation omitted)).

Ex parte Rodriguez, 92 USPQ2d 1395, 1406–07 (BPAI Oct. 1, 2009)
(precedential). That is, the disclosure must teach those skilled in the art how
to make and use the full scope of the claimed invention without undue
experimentation. See Wright, 999 F.2d at 1561.
Here, the disclosure and accompanying references do not provide
adequate information on how to make or use the rules generation step for the
full scope of the invention, which includes any computer-implemented
method “to arbitrate between results to be arbitrated in order to effectuate a
desired result, each of said results to be arbitrated obtained by one of a
plurality of procedures from observationally derived data, each of said
results to be arbitrated having a representative characteristic, said
representative characteristic hereinafter referred to as a type” as recited in
claim 1.
Appeal 2012-003629
Application 11/438,720

11
The disclosure addresses the problem in general in Figure 1 that
shows result characteristics (step 15), results comparisons (step 80), and an
indicator of the degree of success of comparison (step 90) inputted into the
rules generating step (step 20). Yet the disclosure does not address what the
“characteristic” is other than it includes an “indicator” and comparison
results (Spec. ¶ 20) or what the “information derived from said results to be
arbitrated” (Claim 1) is. Thus, we find that undue experimentation is
required by an ordinarily skilled artisan to determine what the
“characteristic” input should be.
Additionally, the disclosure provides an ordinarily skilled artisan with
little information about what the desired output of a general rule-generation
step using a neural network should be other than be “results most likely to be
accurate . . . .” App. Br. 6. Yet, we noted that this discussion addresses the
results of the algorithm and not the rule generation set. Moreover, without
knowing the specific results to be arbitrated, the disclosure provides the
ordinary artisan with little information about the desired results of the rule
generation step. Thus, undue experimentation would be required by an
artisan to determine what the outputs are. Finally, as stated above, undue
experimentation would be required by an ordinarily skilled artisan to
determine what training set to use to generate rules.
Appellants also provide us with a 37 C.F.R. § 1.132 declaration of
Kathleen Chapman (“Chapman Decl.”). See App. Br., Evidence App’x,
E–208–E-214. Ms. Chapman states the disclosure provides enough detail so
an ordinarily skilled artisan could program a computer to perform the
claimed generating rules step without undue experimentation. Chapman
Decl. ¶ 4. However, although Ms. Chapman states she is “a person of
Appeal 2012-003629
Application 11/438,720

12
ordinary skill in the art of Computer Science as evidenced by my
background in the field of Computer Science and Engineering” (Chapman
Decl. ¶ 1), she does not have a master’s degree or greater in computer
science, electrical engineering, or mathematics (Chapman Decl. ¶ 1(i)(a)
AND (b)).6
Appellants further state that Ms. Chapman has eighteen years of
experience as a software engineer. App. Br. 10. Yet, most of the software
or programming experience discussed in the Chapman declaration does not
address artificial intelligence, expert systems, or neural networks techniques.
Chapman Decl. ¶ 1. Paragraph (e) states Ms. Chapman consulted on an “AI
development” project, but this project is listed among numerous others. See
id. Accordingly, Appellants have not demonstrated adequately that she is a
person of ordinary skill in the art
Additionally, Ms. Chapman is a patent attorney that provides
prosecution services for the assignee. Chapman Decl. ¶ 2. Ms. Chapman
also declares that she reviewed claims 6–15 of this application. Chapman
Decl. ¶ 3. Although claim 1 is similar in scope to claims 6 and 11, the
record contains no opinion from Ms. Chapman concerning representative
claim 1. As such, the Chapman declaration has little persuasive weight.
For the first time in the Reply Brief, Appellants discuss in more detail
using a genetic algorithm to generate the rules. Reply Br. 5, 7. Appellants
contend that since the inputs and outputs are “well defined,” a genetic
algorithm can also be used to generate rules. Id. (citing Hisao Ishibuchi et

6 The declaration indicates Ms. Chapman received a bachelor’s degree in
meteorology and minor in mathematics, as well as a master’s degree in
meteorology. Chapman Decl. ¶ 1.
Appeal 2012-003629
Application 11/438,720

13
al., Selecting Fuzzy Rules By Genetic Algorithm for Classification Problems,
2 IEEE Int’l Conf. Fuzzy Sys. 1119–24 (1993) and Takashi Yamamoto &
Hisao Ishibuchi, Performance Evaluation of Three-Objective Genetic Rule
Selections, 2003 IEEE Int’l Conf. on Fuzzy Sys. 149-54 (2003)). However,
for the reasons explained above, we still determine that an ordinarily skilled
artisan would have to experiment unduly to determine the outputs.
Based on the above discussion, we find this factor weighs against the
specification teaching those skilled in the art how to make and use the full
scope of the claimed invention.
(2) Amount of Direction or Guidance Presented
The Examiner finds that there is very little guidance in the disclosure
about how to make or use the rule generation step and that the guidance
provided is generic to rule generation in neural networks and not applicable
to the discussed zip code application. Ans. 13. Appellants argue that how
one generates rules using neural networks is “well-established” and one
skilled in the art would know how to construct and train a neural network to
obtain rules. App. Br. 9 (citing Evidence App’x at E-202–E-214).
Although acknowledging that the disclosure does not need to teach
concepts well-known to an ordinarily skilled artisan (see MPEP
§ 2164.05(a)), we determine that the disclosure at best provides adequate
information to generate rules for the zip code example. However, as stated
above, claim 1 is broader in scope, covering any computer-implemented
method to arbitrate between results to be arbitrated in order to effectuate a
desired results obtained from procedures. App. Br., Claims App’x 1–2.
The Examiner further contends, and we agree, that there is little
guidance for how to make or use this invention for the generic computer-
Appeal 2012-003629
Application 11/438,720

14
implemented method to arbitrate between results to be arbitrated in order to
effectuate desired results obtained from procedures recited in claim 1. See
Ans. 13.
As previously explained, there is little guidance to an ordinarily
skilled artisan concerning what outputs or the training set would be needed
to generate rules using computer techniques, such as neural network.
Granted, Appellants’ Figure 1 addresses the general solution. Spec. ¶ 20;
Fig. 1. However, we are still left to determine what “characteristics” shown
in Figure 1 to input or what “information derived from said results to be
arbitrated” as claimed is inputted to generate rules and a neural network tree.
Moreover, we are not clear on how the training set for a general case would
be obtained.
Lastly, Appellants further state “level of detail is left to the design
engineer working under the guidance of the inventor.” App. Br. 11
(emphasis added); see also Reply Br. 8. Here, Appellants admit that
direction or guidance from the inventor would be needed even in the zip
code example discussed in the disclosure. Thus, the amount of direction in
the general solution of a computer-implemented method to arbitrate between
results is not usual and greater than the routine.
We find this factor weighs against demonstrating the disclosure
teaches those skilled in the art how to make and use the full scope of the
claimed invention.
(3) The Presence or Absence of Working Examples
The Examiner states that the disclosure has no working examples of
the claimed invention. Ans. 13. Appellants contend examples of “how to
quantify rules from experience into an algorithm, similar to an inference
Appeal 2012-003629
Application 11/438,720

15
engine” and “other known approaches to obtain rules, neural networks and
genetic algorithms” have been provided. Reply Br. 8 (citing Spec.
¶¶ 29–31).
Paragraph 29 of the disclosure does not address how the rules are
generated and is directed more towards the inputs to an algorithm at later
step 300. Spec. ¶ 29; Fig. 3. Additionally, we agree with the Examiner that
the examples provided in paragraph 30 are not described as computer-
generated rules and thus are crafted presumably by humans, failing to
discuss “whether rules generated by a processor are intended to be similar in
form to the rules listed in paragraph 30.” Ans. 6. Finally, paragraph 31
discusses various algorithms applied at later step 310 that use the rules in a
particular sequence and not address-generating rules.7
As for paragraph 33, this paragraph addresses generating rules using a
neural networks or a genetic algorithm. Spec. ¶ 33. For the neural network
algorithms, the disclosure provides a citation to a textbook, which has not
been provided or further discussed by Appellants in the briefs.
As such, we agree with the Examiner that the disclosure has no
working examples in the disclosure teaching an ordinarily skilled artisan of
how to make or use the step of “generating rules” using the neural network
or genetic algorithm techniques. Given the above, this factor weighs against

7 We note in passing that paragraphs 30 and 31 have numerous
typographical errors that lead to confusion. For example, paragraph 30
includes rule 1 and then rules 27–30. Spec. ¶ 30. Presumably, rules 27–30
are intended to be rules 2-5. However, the next rule is numbered rule 5 and
then states “[t]he above five rules (2 through 6)[.]” Id. Also, rule 7 states
“[t]his is a variation of rule 7.” Id. Moreover, as the Examiner notes,
paragraph 31 refers to a non-existent rule 21. Spec. ¶ 31.
Appeal 2012-003629
Application 11/438,720

16
demonstrating the disclosure teaches an ordinarily skilled artisan how to
make and use the full scope of the invention.
(4) The Nature of the Invention
The Examiner states that the nature of the invention includes both a
general arbitration and an arbitration of zip code OCR results. Ans. 13–14.
Appellants state that the nature of the invention includes arbitration for
matching arbitrarily derived observational data that leads to rule-based
processing. App. Br. 9. Thus, both the Examiner and Appellants find that
the invention includes a general arbitration technique for arbitrating between
results. We find this factor neither weighs in favor of or against
demonstrating the disclosure teaches an ordinarily skilled artisan how to
make or use the full scope of the invention of claim 1.
(5) The State of the Art
The Examiner finds that the state of prior art includes numerous
examples of generating logic rules, including in a neural network form.
Ans. 14. However, the Examiner argues that the prior art lacks the
application of techniques to arbitration among results in general or zip code
OCR results. Id.
Appellants contend that the state of the art in artificial intelligence is
mature and that there are a number textbooks guiding ordinarily skilled
artisan how to generate rules. App. Br. 5, 9–10 (citing Gallant at 315-30 and
Chapter 14 (not provided)). Additionally, Appellants argue the state of the
art is “that of routine technology, not nascent technology.” App. Br. 9
(citing Evidence App’x at E-202–E-214).
Concerning the citation to galileo-solutions.com, Artificial
Intelligence – State of the Art, available at http://www.galileo-
Appeal 2012-003629
Application 11/438,720

17
solutions.com/2010/08/16/artificial-intelligence-%E2%80%93-state-of-the-
art/ (last visited Aug. 2010), this reference explains that artificial intelligence
is found in different fields, but does not address the topic of the invention —
arbitration for matching arbitrarily derived observational data that leads to
rule-based processing. See App. Br., Evidence App’x, E–204.
The second citation, a website discussing the textbook Neural
Network Learning and Expert Systems by Stephen I. Gallant, available at
http://cognet.mit.edu/library/books/view?isbn=0262071452, addresses
neural networks techniques on a superficial level. See App. Br., Evidence
App’x, E–206. Portions of this textbook have also been provided. See App.
Br., Evidence App’x at E–216–E-269. As addressed above, this provides
some evidence of the state of the art of neural network techniques.
Lastly, when discussing the state of the art, Appellants refer to the
Chapman declaration. See App. Br. 9 (citing Evidence App’x, E–208–E-
214). However, as discussed above, this declaration is not persuasive.
When considering all the evidence, we accept that the state of the art
of neural network techniques is known and mature. Accordingly, this factor
weighs in favor of the disclosure coupled with that what is known by an
ordinarily skilled artisan teaches how to make or use the invention recited in
claim 1.
(6) The Relative Skill of Those in the Art
The Examiner and Appellants agree that the skill in the art is one with
a Master’s degree in or higher in electrical engineering, computer science, or
mathematics. Ans. 14; App. Br. 5, 10. Given that the relative skill of those
in the art is fairly sophisticated, this factor weighs in favor of demonstrating
Appeal 2012-003629
Application 11/438,720

18
the disclosure teaches an ordinarily skilled artisan how to make and use the
claimed invention.
(7) The Predictability or Unpredictability in Art
The Examiner finds that the predictability of this art is dependent on
problem, and as the variables used to generate rules becomes larger, the
neural network becomes more complex and unpredictable. See Ans. 14–15.
Appellants argue that the art is “relatively simple” and thus presumably
predictable. App. Br. 10.
As discussed above, when addressing the experimentation level, we
found that the disclosure provides the ordinary artisan with insufficient
information about how to obtain the desired results or the training set for the
generic “computer implemented method” embodiment that generates rules
as recited in claim 1. On the other hand, we found the state of the art is
mature and the skill level is relatively high. Overall, we find this factor
neither weighs for or against demonstrating the disclosure shows an
ordinarily skilled artisan how to make or use the full scope of claim 1.
(8) The Breadth of the Claims
The Examiner states the claims are fairly narrow. Ans. 15–16.
Appellants indicate the claims are not excessively broad. App. Br. 10.
Although we agree that there are many limitations in claim 1 narrowing its
scope, we highlight the recitation of the “generating rules” is not specific,
other than requiring the rules use as inputs the results to be arbitrated and
some type of “information” derived from said results to be arbitrated.
Additionally, the recited “computer implemented method to arbitrate
between results” is not directed to any particular results or applications.
App. Br., Claims App’x 1–2. As such, we find the breadth of the claims is
Appeal 2012-003629
Application 11/438,720

19
broader than described by either the Examiner or Appellants. On balance,
we find this factor weighs against demonstrating the disclosure shows how
to make or use the full scope of claim 1.
Considering all the Wands factors collectively, we conclude that the
disclosure may teach an ordinarily skilled artisan how to make or use the
claimed “generating rules” step for the specific zip code embodiment when
using a neural network or genetic algorithm technique. On the other hand,
we conclude the disclosure fails to teach how to make or use the same
“generating rules” step for the entire scope of claim 1.
We agree that there is no need to discuss what is well-known to an
ordinarily skilled artisan in the disclosure. App. Br. 7 (citing MPEP
§ 2164.05(a); Reply Br. 5–6). Even so, when considering the known
concepts of generating rules using neural network or genetic algorithm as
stated in the disclosure, we conclude that insufficient information has been
provided to ordinarily skilled artisans for all embodiments encompassed by
the scope of claim 1, such that they would have known how to make or use
the claimed “generating rules” step without undue experimentation.
For the foregoing reasons, Appellants have not persuaded us of error
in the rejection of independent claim 1 and claims 2–15 not separately
argued with particularity.

CONCLUSION
The Examiner did not err in rejecting claims 1–15 under § 112, first
paragraph by finding that the disclosure fails to teach an ordinarily skilled
artisan how to make and use the claimed “generating rules” step without
undue experimentation.
Appeal 2012-003629
Application 11/438,720

20

DECISION
The Examiner’s decision rejecting claims 1–15 is affirmed.
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