Ex Parte Hoya

Board of Patent Appeals and Interferences

Feb 26, 2009

10253642 (B.P.A.I. Feb. 26, 2009)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
____________________

Ex parte TETSUYA HOYA
____________________

Appeal 2008-0024
Application 10/253,642
Technology Center 2100
____________________

Decided:1 February 26, 2009
____________________

Before ALLEN R. MACDONALD, ST. JOHN COURTENAY III, and
THU A. DANG, Administrative Patent Judges.

MACDONALD, Administrative Patent Judge.

DECISION ON APPEAL

1 The two-month time period for filing an appeal or commencing a civil
action, as recited in 37 CFR § 1.304, begins to run from the decided date
shown on this page of the decision. The time period does not run from the
Mail Date (paper delivery) or Notification Data (electronic delivery).

Appeal 2008-0024
Application 10/253,642

STATEMENT OF CASE
Introduction
Appellant appeals under 35 U.S.C. § 134 from a final rejection of
claims 1-32. We have jurisdiction under 35 U.S.C. § 6(b).
According to Appellant, the invention relates to modeling of functions
relating to psychological activities of the brain. More particularly the
invention relates to “a memory system using an artificial neural network
structure (neural network model) for modeling such psychological functions
of the brain as ‘intuition’, ‘consciousness or awareness’, ‘memory-chaining’
and ‘emotion expression’.” (Spec. 1:6-12).

Exemplary Claim(s)
Exemplary independent claim 1 under appeal reads as follows:
1. A memory system adapted to model psychological functions of a
brain by using an artificial neural network structure, comprising:

a short-term memory neural network unit formed that temporarily
stores input pattern vectors: and

a long-term memory neural network unit formed on the basis of
output vectors provided by the short-term memory neural network unit;

wherein the long-term memory neural network unit includes
hierarchical neural networks for ordinary outputs, formed by using the
output vectors of the short-term memory neural network unit.

2
Appeal 2008-0024
Application 10/253,642

Prior Art
The prior art relied upon by the Examiner in rejecting the claims on
appeal is:
Graupe US 5,920,852 July 6, 1999
Passera US 6,347,310 B1 Feb. 12, 2002
Sadakuni US 6,446,056 B1 Sep. 3, 2002
Rejections
The Examiner rejected claims 1-3, 5, and 7-11 under 35 U.S.C.
§ 102(b) as being anticipated by Graupe.
The Examiner rejected claims 4, 6, 11-17, 25, 26, 30, and 31 under
35 U.S.C. § 103(a) as being unpatentable over the combination of Graupe
and Passera.
The Examiner rejected claims 18-24 and 27-29 under 35 U.S.C.
§ 103(a) as being unpatentable over the combination of Graupe, Passera, and
Sadakuni.
The Examiner rejected claims 1-32 under 35 U.S.C. § 101 as being
directed to non-statutory subject matter.
Appellant’s Contentions
(1) Appellants contend (App. Br. 7) that the subject matter of claims
1-3, 5, and 7-11 is not anticipated because the Examiner erred in finding
that Graupe teaches “the long-term memory neural network unit includes
hierarchical neural networks for ordinary outputs, formed by using the
output vectors of the short-term memory neural network unit” (Ex. Ans. 5).
(2) Appellants contend (App. Br. 12) that the subject matter of claims
4, 6, 11-17, 25, 26, 30, and 31 is not obvious (a) because the Examiner erred
in finding that Passera teaches “the neurons of the long-term memory
3
Appeal 2008-0024
Application 10/253,642

network unit has a radial-basis function main part that provides an activation
intensity corresponding to similarity between an input vector and a centroid
vector according to a radial-basis function” (Ex. Ans. 11), and (b) because
the Examiner erred as discussed in (1) above.
(3) Appellants contend (App. Br. 14) that the subject matter of claims
18-24 and 27-29 is not obvious because the Examiner erred as discussed in
(1) and (2) above.
(4) Appellant contends that the Examiner erred in rejecting claims
1-32 under 35 U.S.C. § 101 as failing to recite statutory subject matter
because “there is not a separate ‘technological arts” test for determining
statutory subject matter under 35 U.S.C. § 101.” (App. Br. 15).
(5) Appellant also contends that the Examiner erred in rejecting
claims 1-20 under 35 U.S.C. § 101 because “the Examiner’s [assertion] that
none of the claims are drawn to structural limitations or a tangible device is
incorrect.” (App. Br. 15).
Result
We affirm.

ISSUE(S)
Issues on Appeal
(1)
Whether Appellant has shown that the Examiner has erred in rejecting
claims 1-3, 5, and 7-11 under 35 U.S.C. § 102(b) because Graupe does not
describe limitations required by these claims?
4
Appeal 2008-0024
Application 10/253,642

(2)
Whether Appellant has shown that the Examiner has erred in rejecting
claims 4, 6, and 11-31 under 35 U.S.C. § 103(a) because Passera does not
demonstrate that limitations required by these claims are known in the prior
art?
(3)
Whether Appellant has shown that the Examiner has erred because the
Examiner has made an invalid “technological arts” rejection under
35 U.S.C. § 101?
(4)
Whether Appellant has shown that the Examiner has erred because the
construction of claims 1-32 requires that the claimed inventions be construed
as tangible?

FINDINGS OF FACT
The following Findings of Fact (FF) are shown by a preponderance of
the evidence.
Appellant’s Invention
1. Appellant invention relates to “a memory system using an
artificial neural network structure (neural network model) for modeling such
psychological functions of the brain as ‘intuition’, ‘consciousness or
awareness’, ‘memory-chaining’ and ‘emotion expression’.” (Spec. 1:7-12).
2. An object of the present invention is “to provide a memory
system (‘memory-chaining’ system) adapted to realize ‘memory-chaining’
and ‘emotion expression’ by using a comparatively simple artificial neural
5
Appeal 2008-0024
Application 10/253,642

network structure, a memory-chaining program product, and a neuron
element for use in the memory chaining system.” (Spec. 2:21-26).
3. The term “memory-chaining” signifies “an act of chaining
information or processes held for long-term memory, and, psychologically,
corresponds to memory called episodic memory, declarative memory or
procedural memory.” (Spec. 2:26-30).
4. The present invention proposes, as a basis for modeling the
psychological functions of the brain, a hierarchically arranged generalized
regression neural network (HA-GRNN) including a hierarchical memory
system provided with two memory models, a short-term memory (STM)
model and a long-term memory (LTM) model, and interprets and embodies
‘intuition’, the state of ‘awareness’, ‘memory-chaining’ and ‘emotion
expression’ in terms of the evolution of the HA-GRNN. (Spec. 2:32
through 3:2).
5. The configuration of the memory system in the first
embodiment of the present invention is shown by FIGS. 1 to 4
(Spec. 16:23-25) as follows:
Referring to FIG. 1, a memory system 10 in the first
embodiment of the present invention is a hierarchical memory
system including a short-term memory (STM) model and a
long-term memory (LTM) model. The memory system 10
includes a STM network (short-term memory neural network
unit) 11 for realizing the STM model; a LTM network group
(LTM networks 1 to L) 12 for realizing the LTM model; and a
decision unit 13 of a winner-take-all system. The LTM network
group 12 and the decision unit 13 constitute a hierarchical LTM
network unit (long-term memory neural network unit) 21.

6
Appeal 2008-0024
Application 10/253,642

Shown in FIG. 1 are an input pattern vector x applied to
the memory system 10, an output vector OSTM provided by the
STM network 11, outputs OLTMi (i=1, 2, . . . , L) provided by the
LTM networks 1 to L of the LTM network group 12, weights vi
for weighting the outputs of the LTM network group 12, and an
output ONET (perceptional output) of the memory system 10.
The LTM network group 12 is subdivided into a first part
including the LTM network 1 for providing ‘intuitive outputs’,
and a second part including the LTM networks 2 to L for
providing ‘ordinary outputs’.
Whereas the LTM networks 2 to L are the same in
configuration as the GRNN shown in FIG. 2B, the STM
network 11 and the LTM network 1 have a configuration
different from that of the ordinary GRNN shown in FIG. 2B
and are modifications of a RBF-NN as shown in FIGS. 3 and 4,
respectively.
The respective configurations of the STM network 11
and the LTM networks 1 to L will be described with reference
to FIGS. 2B, 3 and 4.
(Spec. 16:26- through 17:16).
6. The LTM network 1 is formed after the long, iterative
exposition of input pattern vectors that highly activate a particular number of
centroids in the LTM networks 2 to L. In other words, the transition of the
centroids from the STM network to the LTM networks 2 to L is regarded as
a normal learning process, and the transition of the centroids from the LTM
networks 2 to L to the LTM network 1 provides a chance for the memory
system 10 to generate ‘intuitive’ outputs. (Spec. 23:22-29).
7. A method of forming an artificial neural network structure
modeling the psychological functions of the brain in the memory system 10
in the first embodiment can be easily realized in a program product that can
be executed by a computer system 40 as shown in FIG. 13. The computer
7
Appeal 2008-0024
Application 10/253,642

system 40 includes a processor 41, a memory 42, a hard disk 43, and
peripheral devices including an input device 44 including a keyboard and a
mouse, output device 45 including a display and a printer, a flexible disk
drive (FD drive) 46, a CD-ROM drive 47, and a bus 48 interconnecting
those component devices. The program is stored in a readable recording
medium, such as a FD 49 or a CD-ROM 50; and the processor 41 reads
instructions included in the program sequentially from the recording
medium, and executes the instructions to carry out the foregoing procedures.
(Spec. 27:26 through 28:3).
Appellant’s Admissions
8. A generalized regression neural network (GRNN), a
prerequisite for the HA-GRNN, is one of the theories of artificial neural
networks in existence today and falls in the category of radial-basis function
neural networks (RBF-NNs). However, the GRNN, unlike ordinary
RBF-NNs, has a special property that the weight vectors between radial-
basis functions (RBFS) and output neurons are given identical with target
vectors. By virtue of this attractive property, a dynamic neural system can be
modeled without requiring any complex mathematical operations.
(Spec. 3:3-12).
9. The invention of independent claim 1 is shown at page 16,
line 23 through page 17, line 16, of Appellant’s specification. (App. Br. 3).
(See FF 5 above).

PRINCIPLES OF LAW
Appellant has the burden on appeal to the Board to demonstrate error
in the Examiner’s position. See In re Kahn, 441 F.3d 977, 985-86 (Fed. Cir.
8
Appeal 2008-0024
Application 10/253,642

2006) (“On appeal to the Board, an applicant can overcome a rejection
[under § 103] by showing insufficient evidence of prima facie obviousness
or by rebutting the prima facie case with evidence of secondary indicia of
nonobviousness.”) (quoting In re Rouffet, 149 F.3d 1350, 1355 (Fed. Cir.
1998)). See also Hyatt v. Dudas, 492 F.3d 1365, 1369 (Fed. Cir. 2007) (“As
we explained in In re Oetiker, the prima facie case is merely a procedure
device that enables an appropriate shift of the burden of production.”) See
Id. at 1369-70 (“Once the applicant is so notified, the burden shifts to the
applicant to rebut the prima facie case with evidence and/or argument.”)
For a rejection under § 102, Appellant may sustain this burden by
showing that the prior art reference relied upon by the Examiner fails to
disclose an element of the claim. It is axiomatic that anticipation of a claim
under § 102 can be found only if the prior art reference discloses every
element of the claim. See In re King, 801 F.2d 1324, 1326 (Fed. Cir. 1986)
and Lindemann Maschinenfabrik GMBH v. American Hoist & Derrick Co.,
730 F.2d 1452, 1457 (Fed. Cir. 1984).
“Section 103 forbids issuance of a patent when ‘the differences
between the subject matter sought to be patented and the prior art are such
that the subject matter as a whole would have been obvious at the time the
invention was made to a person having ordinary skill in the art to which said
subject matter pertains.’” KSR Int'l Co. v. Teleflex Inc., 127 S. Ct. 1727,
1734 (2007). The question of obviousness is resolved on the basis of
underlying factual determinations including (1) the scope and content of the
prior art, (2) any differences between the claimed subject matter and the
prior art, (3) the level of skill in the art, and (4) where in evidence, so-called
9
Appeal 2008-0024
Application 10/253,642

secondary considerations. Graham v. John Deere Co., 383 U.S. 1, 17-18
(1966). See also KSR, 127 S. Ct. at 1734 (“While the sequence of these
questions might be reordered in any particular case, the [Graham] factors
continue to define the inquiry that controls.”)
Under [the Graham] framework, once an Examiner demonstrates that
the elements are known in the prior art and that one of ordinary skill could
combine the elements as claimed by known methods and would recognize
that the capabilities or functions of the combination are predictable, then the
Examiner has made a prima facie case that the claimed subject matter is
likely to be obvious. The burden then shifts to the Appellant to show that
the Examiner erred in these findings or to provide other evidence to show
that the claimed subject matter would have been nonobvious.
“[R]ejections on obviousness grounds cannot be sustained by mere
conclusory statements; instead, there must be some articulated reasoning
with some rational underpinning to support the legal conclusion of
obviousness.” KSR, 127 S. Ct. at 1741 (citing In re Kahn, 441 F.3d 977, 988
(Fed. Cir. 2006)).
Claim Construction
"Our analysis begins with construing the claim limitations at issue."
Ex Parte Filatov, No. 2006-1160, 2007 WL 1317144, at *2 (BPAI 2007).
"The Patent and Trademark Office (PTO) must consider all claim
limitations when determining patentability of an invention over the prior
art." In re Lowry, 32 F.3d 1579, 1582 (Fed. Cir. 1994) (citing In re Gulack,
703 F.2d 1381, 1385 (Fed. Cir. 1983)). "Claims must be read in view of the
specification, of which they are a part." Markman v. Westview Instruments,
10
Appeal 2008-0024
Application 10/253,642

Inc., 52 F.3d 967, 979 (Fed. Cir. 1995) (en banc). "[T]he PTO gives claims
their 'broadest reasonable interpretation.'" In re Bigio, 381 F.3d 1320, 1324
(Fed. Cir. 2004) (quoting In re Hyatt, 211 F.3d 1367, 1372 (Fed. Cir. 2000)).
"Moreover, limitations are not to be read into the claims from the
specification." In re Van Geuns, 988 F.2d 1181, 1184 (Fed. Cir. 1993)
(citing In re Zletz, 893 F.2d 319, 321 (Fed. Cir. 1989)).

ANALYSIS
§ 102 - Claims 1-3, 5, and 7-11
Appellant argues that the Examiner’s finding that Graupe teaches “the
long-term memory neural network unit includes hierarchical neural networks
for ordinary outputs, formed by using the output vectors of the short-term
memory neural network unit” (Ex. Ans. 5) is erroneous. Specifically,
Appellant argues:
[N]one of the SOM modules of Graupe are formed on the basis
of output vectors provided by a short-term SOM module. The
Examiner points to the correlation link Lij of Fig. 1 as being
equivalent to the claimed output vector, however, these
correlation links are not equivalent to an output vector of a
neural network unit. To contrary, the correlation links are
statistical correlations between the various concepts/
words/patterns W1, W2, W3 within the SOM modules. One
skilled in the art would readily appreciate that these correlation
links are not equivalent to the claimed output vectors.
(App. Br. 7).
We agree. We do not find the argued limitation in Graupe. The
Appellant has shown error in the Examiner’s prima facie case.
For this reason, Appellant has established that the Examiner erred
with respect to this rejection of claims 1-3, 5, and 7-11 under § 102.
11
Appeal 2008-0024
Application 10/253,642

§ 103 - Claims 4, 6, 11-17, 25, 26, 30, and 31
Appellant argues that the Examiner’s finding that Passera teaches “the
neurons of the long-term memory network unit has a radial-basis function
main part that provides an activation intensity corresponding to similarity
between an input vector and a centroid vector according to a radial-basis
function” (Ex. Ans. 11) is erroneous. Specifically, Appellant argues that
“[n]owhere in the cited passage [Passera at col. 5, ll. 5-11] is there any
disclosure or suggestion of a radial basis function, much less providing an
activation intensity . . . as claimed.” (App. Br. 12).
We agree. The Graham framework requires that the Examiner
demonstrates that the claim elements are known in the prior art in order to
make a prima facie case that the claimed subject matter is likely to be
obvious. We do not find the argued claim element in the cited passage. The
Appellant has shown error in the Examiner’s prima facie case.
Therefore, for this reason and the reason discussed above with respect
to the § 102 rejection, Appellant has established that the Examiner erred
with respect to this rejection of claims 4, 6, 11-17, 25, 26, 30, and 31 under
§ 103(a).
§ 103 - Claims 18-24 and 27-29
For the reasons above, Appellant has also established that the
Examiner erred with respect to this rejection of claims 18-24 and 27-29
under § 103(a).
12
Appeal 2008-0024
Application 10/253,642

§ 101 - Claims 1-32
Appellant argues the Examiner erred in rejecting claims 1-32 under
35 U.S.C. § 101 because there is not a separate “technological arts” test for
determining statutory subject matter.
We disagree. While we agree that there is no “technological arts” test,
we do not find that the Examiner relies on a “technological arts” test.
Rather, the Examiner summarized that the claims are not statutory because
they do not require a tangible device or a method of using such a device or
producing real-world results. We find that the essence of the Examiner’s
summary conforms to the requirements set forth in In re Bilski, 545 F.3d 943
(Fed. Cir. 2008) and In re Nuijten, 500 F.3d 1346 (Fed. Cir. 2007).
Appellant also contends that the Examiner’s tangibility summary is
incorrect. However, the Appellant only presents an argument as to claim 1
(which argument is also applicable to claims 2-4 and 12-24). No argument
on this point is presented as to the remaining claims (the remaining claims
do not share the particular “artificial neural network unit” feature argued
with regard to claim 1). Therefore, Appellant has waived any such
arguments with respect to these claims.
As to claim 1, Appellant argues that “[o]ne skilled in the art would
readily appreciate that an artificial neural network unit refers to a computer
architecture in which processors are connected in a specified manner.” (App.
Br. 15:17-19). We disagree. Appellant points to no evidence in the record
to support this argued definition is the required definition of an artificial
neural network unit. To the contrary, Appellant’s own tangible embodiment
of an artificial neural network unit (FF 7) does not comport to this definition
13
Appeal 2008-0024
Application 10/253,642

as it contains only a single processor. The real issue raised by the Examiner
is whether the term “artificial neural network unit” also encompasses an
intangible embodiment consisting of a mere data structure to be
implemented on the single processor. Appellant asserts that the invention of
claim 1 is shown at page 16, line 23 through page 17, line 16, of Appellant’s
Specification. (FF 9). The elements of claim 1 shown in this section of
Appellant’s Specification are described as models, networks, or units, which
an artisan would appreciate could refer either to tangible structures as
Appellant argues, or could refer to the intangible data structures
underpinning the tangible structures. We conclude that the Examiner has
correctly concluded that claim 1 is not required to be the tangible
embodiment.
Because we have concluded that claim 1 covers at least one
embodiment which is directed to subject matter that is unpatentable under
§ 101, we further conclude that claim 1 is unpatentable as being directed to
non-statutory subject matter.2 Cf. Amgen v. Hoechst Marion Roussel, 314
F.3d 1313, 1329 (Fed. Cir. 2003); see also MPEP § 2105.
Therefore, for the reasons above, Appellant has not established that
the Examiner erred with respect to this rejection of claims 1-32 under § 101.
We note that Appellant is not precluded from amending claims 1-32,
in a continuing application, so that the claims are limited to only the tangible
embodiments.

2 “If a claim covers material not found in any of the four statutory
categories, that claim falls outside the plainly expressed scope of § 101 even
if the subject matter is otherwise new and useful.” In re Nuijten, 500 F.3d
1346, 1354 (Fed. Cir. 2007).
14
Appeal 2008-0024
Application 10/253,642

CONCLUSION OF LAW
(1) Appellant has established that the Examiner erred in rejecting
claims 1-3, 5, and 7-11 as being unpatentable under 35 U.S.C. § 102(b).
(2) Appellant has established that the Examiner erred in rejecting
claims 4, 6, and 11-31 as being unpatentable under 35 U.S.C. § 103(a).
(3) Appellant has failed to establish that the Examiner erred in
rejecting claims 1-32 as being unpatentable under 35 U.S.C. § 101.
(4) Claims 1-32 are not patentable.

DECISION
The Examiner's rejection of claims 1-32 under 35 U.S.C. § 101 is
affirmed.
The Examiner's rejection of claims 1-3, 5, and 7-11 under
35 U.S.C. § 102(b) is reversed.
The Examiner's rejections of claims 4, 6, and 11-31 under
35 U.S.C. § 103(a) are reversed.
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

pgc

BIRCH STEWART KOLASCH & BIRCH
PO BOX 747
FALLS CHURCH VA 22040-0747
15