Ex parte MOLL et al.

Board of Patent Appeals and Interferences

Jun 11, 1998

08202609 (B.P.A.I. Jun. 11, 1998)

Application for patent filed February 28, 1994. According1
to the appellants, this application is a continuation of
Application 07/783,016, filed October 25, 1991, now U.S. Patent
No. 5,355,313, issued October 11, 1994.
1
THIS OPINION WAS NOT WRITTEN FOR PUBLICATION
The opinion in support of the decision being entered today (1)
was not written for publication in a law journal and (2) is not
binding precedent of the Board.
Paper No. 18
UNITED STATES PATENT AND TRADEMARK OFFICE
_____________
BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
_____________
Ex parte ROBERT F. MOLL, WILLIAM C. PEARSON,
JOHN R. ROGERS and
JACKY M. WIENER
_____________
Appeal No. 97-0507
Application 08/202,6091
______________
ON BRIEF
_______________
Before THOMAS, BARRETT and CARMICHAEL, Administrative Patent
Judges.
THOMAS, Administrative Patent Judge.
Appeal No. 97-0507
Application 08/202,609
2
DECISION ON APPEAL
Appellants have appealed to the Board from the examiner’s
final rejection of claims 25 to 30, which constitute all the
claims remaining in the application.
Claim 25 is reproduced below:
25. A system for determining the existence of subsurface
earth structural anomalies using aeromagnetically measured date
representative of magnetic field strength at known positions
above the surface of the earth as input data, and for producing
processed output data comprising plots of depth to basement rock
at corresponding known positions above the surface of the earth,
said system comprising:
a programmed digital computer having a memory and forming a
neural network having a multiple element input layer each element
of which is capable of receiving input digital data comprising
representations of magnetic field strengths at known positions
above the surface of the earth and for producing output signals,
each element of which is connected to plural elements of a hidden
layer of elements via a first weighted system or
interconnections, the weights of which are adjustable under
program control, and said hidden layer elements each having a sum
function and a transfer function associated therewith for summing
all input signals to each element and for applying said transfer
function to such sum to produce an output signal from each
element, and an output layer of elements each of which has an
input connected to pliral hidden layer elements via a second
weighted system of interconnections, the weights of which are
adjustable under program control, said output layer elements each
having a sum function and a transfer function associated
therewith for summing all input signals to each element and for
applying said transfer function to such sum to produce an output
signal representative of depth to basement rock at known
positions above the surface of the earth; and
program means stored in said computer memory for training
said programmed digital computer forming a neural network by
Appeal No. 97-0507
Application 08/202,609
A rejection of certain claims under 35 U.S.C. § 102(b) as2
set forth in the final rejection has not been maintained or
repeated in the answer. Therefore, it is not before us.
3
repetitively inputting representative aeromagnetically measured
data above the surface of the earth over known depth to basement
rock regions and for repetitively comparing the output of said
output layer elements with said known depth to basement rock to
form error signals, and for changing as a function of said error
signals the weighting functions of said first and second systems
of interconnections between said input layer and said hidden
layer and said output layer of elements so as to minimize the
magnitude of said error signals, until such magnitude is smaller
than a predetermined accuracy value.
The following references are relied on by the examiner:
McCormack 5,265,192 Nov. 23, 1993
(filed Sep. 20, 1990)
Reilly et al. (Reilly), “An Overview of Neural Networks: Early
Models to Real World Systems,” An Intro. To Neural and Electronic
Networks, Academic Press, pp. 227-248 (1990).
Claims 25 to 30 stand rejected under 35 U.S.C. § 102(a) as
being clearly anticipated by Reilly. These same claims also
stand rejected under 35 U.S.C. § 102(e) as being clearly
anticipated by McCormack. 2
Rather than repeat the positions of the appellants and the
examiner, reference is made to the briefs and the answer for the
respective details thereof.
OPINION
At the outset, we note that appellants indicate at the top
of page 3 of the brief that a companion divisional application
Appeal No. 97-0507
Application 08/202,609
4
having the same effective filing date as the present application
“has now issued as U.S. Patent 5,355,313 and contains claims to
the invention in method format.” In contrast, the present claims
on appeal are all apparatus claims.
For the reasons set forth by the examiner in the answer, we
will sustain the rejection of claims 25 to 30 under 35 U.S.C. §
102 as being clearly anticipated by Reilly and the separate
rejection of the same claims under 35 U.S.C. § l02 as being
clearly anticipated by McCormack. We do so even though we
conclude that the more persuasive argumentative approach for the
examiner to have taken on the facts presented in this case would
have been to reject the claims under 35 U.S.C. § 103 in each
instance since it clearly would have been obvious for the artisan
to have programmed a neural network to process any type of data
whether it be aeromagnetic input data, the seismic trace-type of
data as in McCormack or generic types of data to solve real world
problems in a particular environment as in Reilly’s INTRODUCTION.
In an analogous manner, the nature of neural networks is such
that they are enabled to process all types of data, inherently
within 35 U.S.C. § 102, even the claimed aeromagnetic data.
Contrary to appellants’ initial position in the brief, the
examiner does not appear to us to dissect representative
Appeal No. 97-0507
Application 08/202,609
5
independent claim 25 by disregarding the preamble. A brief study
of independent claim 25 yields that the features relating to
aeromagnetic data in the preamble find corresponding language
utilized in the body of this claim even though the body
recitations do not utilize the word “said” to respectively and
clearly refer back to the preamble recitations. The plots in the
preamble are not produced by the system in the body of the claim,
however.
Appellants do not argue any structural distinction per se in
apparatus claim 25 over that which the examiner has argued to
exist in each of the references relied upon under 35 U.S.C.
§ 102. Indeed, appellants’ disclosure of the base neural network
structure in Fig. 1 of the disclosure is shown in the respective
portions of the references relied on by the examiner, namely,
Fig. 11 of Reilly at page 234 of this reference and Figs. 3 and 4
of McCormack. Each reference relates to the basic neural network
structure having an input, hidden and output layers, which are
essentially recited in the initial long clause of the body of
claim 25 on appeal.
Additionally, each reference relied upon teaches training in
the form of backward propagation of error signals based upon
known data through an experimentally determined number of
Appeal No. 97-0507
Application 08/202,609
6
iterations to reach a learned or trained state, as recited in the
second clause of the body of claim 25 on appeal. Indeed, this
portion of claim 25 is based upon appellants’ specification
statement at the top of page 8 that for the backpropagation
chosen in the disclosed invention “the well known Delta Rule is
used for weight adjustment during learning.” This same rule is
discussed at col. 7, beginning at line 31 of McCormack as
referenced by the examiner in the answer. We also observe that
aeromagnetic data has been recognized by appellants’ statement of
the prior art of the invention at page 1 and 2 of the
specification as filed to have existed in the prior art.
The program means (second) clause of the body of claim 25
relating to the training operation of the programmed computer
utilized to simulate a neural network (first clause of the body
of claim 25) is stated to repetitively input representative
aeromagnetic data above the surface of the earth over known depth
to basement rock regions. This says nothing more than a neural
network needs to be trained for processing new data based upon
known data. Both references teach this and also teach that the
weighting functions of the neural network itself are adjusted, as
claimed, during this training operation.
Appeal No. 97-0507
Application 08/202,609
7
Therefore, we are in agreement with the examiner’s
(unedited) position set forth at pages 5 and 6 of the answer that
we reproduce here:
Although Reilly and Cooper publication
does not explicitly disclose that their
neural network may be used for the purpose of
analyzing aeromagnetic data, using specific
data (here, aeromagnetic data) in data
processing or training of a neural network
does not amount to any structural or
functional limitation of the claimed neural
network. The preamble of claim 25 and the
elements (1), (2) and (3) as stated on page 4
of the Appeal Brief, only limit the type of
data used in the claimed neural network. The
specific data used in a neural network does
not impart structure or function beyond the
functional steps normally followed in back
propagation training of a neural network; the
steps are the same regardless of data. That
is, changing the nature of data does not
change the functionality of a neural network
in terms of steps of data processing, error
back propagation and learning. Nor does
changing the data changes the structure,
architecture, and interconnection of a neural
network data processing system. Therefore,
Reilly and Cooper publication discloses each
and every structural limitation of the neural
network as claimed.
This reasoning is also repeated by the examiner at pages 6 and 7
of the answer as to the McCormack patent.
The above positions of the examiner recognize that both
references relied on do not explicitly disclose that the
respective neural networks may be used for the purpose of
Appeal No. 97-0507
Application 08/202,609
8
analyzing aeromagnetic data. Rather than disregarding or
dissecting out such limitations the examiner has merely not given
them patentable weight even within 35 U.S.C. § 102 because they
are statements of intended use. There is adequate precedent for
the examiner’s view that using specific data in the processing of
data and training of a neural network does not amount to any
structural or functional limitation of the claimed neural
network. A different intended use of the same structure as in
the prior art does not prohibit a statutory anticipation
rejection. Indeed, it has been stated by our reviewing court
that “the absence of a disclosure relating to
function does not defeat the Board’s finding of anticipation. It
is well settled that the recitation of a new intended use for an
old product does not make a claim to that old product patentable
(case citations omitted).” In re Schreiber, 128 F.3d 1473, 1477,
44 USPQ2d 1429, 1431 (Fed. Cir. 1997). The court concludes at
128 F.3d 1477, 44 USPQ2d 1431-32, that “Schreiber’s contention
that his structure will be used to dispense popcorn does not have
patentable weight if the structure is already known, regardless
of whether it has ever been used in any way in connection with
Appeal No. 97-0507
Application 08/202,609
9
popcorn” (emphasis added). This reasoning also confirms the
proprietary of the examiner’s approach not to give patentable
weight to the aeromagnetic data input into the neural network
claimed.
The answer does not rely upon Ex parte Masham, 2 USPQ2d 1647
(Bd. Pat. App. & Int. 1987), but the final rejection does make
reference to it at the top of page 2. The examiner made
reference there to this prior Board’s decision indicating that
this case held that a recitation with respect to the manner in
which a claimed apparatus is intended to be employed does not
differentiate the claimed apparatus from a prior art apparatus
satisfying the claimed structural limitations. Schreiber
confirms this. Appellants’ rebuttal of the examiner’s reliance
upon Masham at page 5 of the principal Brief on appeal is
misplaced. That the board affirmed a rejection of a claim based
upon a patent having the same utility as that of the claim on
appeal before it was not dispositive. What appears to us to have
been dispositive in that case was that the recitation with
respect to a different material intended to be worked upon by the
claimed apparatus did not impose any structural limitation upon
Appeal No. 97-0507
Application 08/202,609
10
the claimed apparatus to differentiate it from prior art
apparatus satisfying structural limitations of that which have
been recited on the claims on appeal. In an analogous manner
here, we and the examiner take the view that the nature of the
data to be processed by the neural network is not dispositive or
does not differentiate a prior art neural network apparatus
processing different types of raw data, and particularly data
that has been known and admitted to be known in the prior art by
appellants here. Id. 2 USPQ2d at 1648. Accord, Ex parte
Wikdahl, 10 USPQ2d 1546, 1548 (Bd. Pat. App. & Int. 1989) and In
re Casey, 370 F.2d 576, 580, 152 USPQ 235, 238 (CCPA 1967).
Appellants’ positions set forth at page 2 of the reply brief
are also not persuasive. We are not persuaded by appellants’
reasoning that the structure of the neural network is changed
when the weighting functions are changed as asserted here. As
the examiner’s reasoning makes clear, the references relied on
make clear as well as appellants’ own admission with respect to
prior art neural networks makes clear, the backpropagation
“trainability” of a neural network inherently requires that the
weighting functions be changed in accordance with the training
Appeal No. 97-0507
Application 08/202,609
11
operation associated with known data before the neural network
essentially is “turned loose” upon raw, new data. Changing the
weighting functions associated with a training operation of a
neural operation does not change the structure or the hardware of
the associated electronics of the actual device. The electrical
values do change but the actual electrical circuits processing
the trainable data values do not change. The mathematical values
or weights change during training in any neural network that is
“trainable.” The act of changing such weighting functions occurs
inherently in any neural network that is “trainable.” The
specific weighting functions changed in accordance with the last
clause of the body of claim 25 on appeal are not recited in the
claim to be changed from one form to another. It is inherent
within the nature of the devices of the prior art relied upon by
the examiner that, for example, the seismic data traces processed
by McCormack may have yielded identical weighting functions to
those only generically recited in this clause of claim 25 on
appeal. The “trainability” recited in the last clause of the
body of claim 25 on appeal affecting the sum function and
transfer function recited in two instances in the first clause of
the body of claim 25 on appeal are inherent properties of the
Appeal No. 97-0507
Application 08/202,609
12
prior art structure in each reference relied upon by the examiner
to reject the claims on appeal.
To further buttress all these considerations, the programmed
digital computer clause is merely “capable of” receiving and
producing specified data, and the program means clause is merely
“for training” when aeromagnetic data is presented. Neither
clause of the body of claim 25 on appeal positively recites the
respective acts or functions are carried out by the “system” of
the preamble. The chosen language refers to future acts or
functions which may or may not occur.
Since appellants’ briefs do not indicate any grouping for
the claims on appeal and because there are no arguments presented
as to dependent claims 26 through 30 as to each rejection of the
claims on appeal, they all fall with our affirmance of both
rejections of claim 25. In view of the foregoing, we have
sustained both rejections of claim 25 to 30 under 35 U.S.C.
§ 102. Therefore, the decision of the examiner is affirmed.
No time period for taking any subsequent action in
connection with this appeal may be extended under 37 CFR §
1.136(a).
Appeal No. 97-0507
Application 08/202,609
13
AFFIRMED
)
JAMES D. THOMAS )
Administrative Patent Judge )
)
)
) BOARD OF PATENT
LEE E. BARRETT )
Administrative Patent Judge ) APPEALS AND
)
) INTERFERENCES
)
JAMES T. CARMICHAEL )
Administrative Patent Judge )
Kenneth R. Priem
Intellectual Property
TEXACO INC.
Appeal No. 97-0507
Application 08/202,609
14
P. O. Box 430
Bellaire, TX 77402-0430