Ex Parte Lakshminarayan et al

Patent Trial and Appeal Board

Sep 22, 2016

13157009 (P.T.A.B. Sep. 22, 2016)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR
13/157,009 06/09/2011 Choudur Lakshminarayan
56436 7590 09/26/2016
Hewlett Packard Enterprise
3404 E. Harmony Road
Mail Stop 79
Fort Collins, CO 80528
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
ATTORNEY DOCKET NO. CONFIRMATION NO.
82690114 6235
EXAMINER
SUGENT, JAMES F
ART UNIT PAPER NUMBER
2175
NOTIFICATION DATE DELIVERY MODE
09/26/2016 ELECTRONIC
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.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address( es):
hpe.ip.mail@hpe.com
mkraft@hpe.com
chris.mania@hpe.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte CHOUDUR LAKSHMINARA YAN, ALEXANDER SINGH
ALVARADO, and JOSE C. PRINCIPE
Appeal2014-009124
Application 13/157,009
Technology Center 2100
Before ERIC S. FRAHM, LARRY J. HUME, and JOHN D. HAMANN,
Administrative Patent Judges.
FRAHM, Administrative Patent Judge.
DECISION ON APPEAL
STATEMENT OF CASE
Introduction
Appellants appeal under 35 U.S.C. § 134(a) from a final rejection of
claims 1-20. We have jurisdiction under 35 U.S.C. § 6(b ).
We affirm.
Disclosed and Claimed Invention
Appellants' disclosed invention relates to adaptive sampling of
integrate and fire (IF) time encoded neural (i.e., brain) signals (Fig. 1; Spec.
i-fi-11-3 and 12-14; Abs.) in order to separate neurons into different classes
(Spec. i-fi-1 17, 25, 26, 31; Fig. 2 shows class assignment module 220
Appeal2014-009124
Application 13/157,009
operating as described in if 31; Fig. 5 shows classes N 1 and N2 of neurons
plotted in Figs. 4a and 4b as described in iii! 33-36).
Exemplary Claim
Exemplary independent claim 1 under appeal, with emphasis added to
the contested limitation, reads as follows:
1. A method of time encoding using an integrate and fire
(IF) sampler, comprising:
receiving input signals for separate classes;
generating a pulse train based on the input signals; and
binning the pulse train to generate a feature vector for
each of the separate classes.
Examiner's Rejections
( 1) The Examiner rejected claims 1--4, 7, 11-14, 1 7, and 18 as being
unpatentable under 35 U.S.C. § 103(a) over the combination of Alvarado,
Alexander Singh et al., Stimulus reconstruction from the biphasic integrate-
and-fire sampler, PROCEEDINGS OF THE 4rn INTERNATIONAL IEEE EMBS
CONFERENCE ON NEURAL ENGINEERING ANTAL YA, TURKEY, pp. 415-18
(April 29 - May 2, 2009) (hereinafter, "Alvarado") and Nedungadi, Aatira et
al., Analyzing multiple spike trains with nonparametric granger causality, J
COMPUT NEUROSCI, Vol. 27, pp. 55---64 (Jan. 10, 2009) (hereinafter,
"Nedungadi"). Final Act. 3-5; Ans. 3---6.
(2) The Examiner rejected claims 5, 6, 15, and 19 as being
unpatentable under 35 U.S.C. § 103(a) over the combination of Alvarado,
Nedungadi, and Sanchez, Justin C., et al., Technology and Signal Processing
2
Appeal2014-009124
Application 13/157,009
for Brain-Machine interfaces, IEEE SIGNAL PROCESSING MAGAZINE, pp. 29-
40 (2008) (hereinafter, "Sanchez"). Final Act. 5---6; Ans. 6-7.
(3) The Examiner rejected claims 8, 9, 16, and 20 as being
unpatentable under 35 U.S.C. § 103(a) over the combination of Alvarado,
Nedungadi, and Lansky, Petr et al., Classification of stimuli based on
stimulus-response curves and their variability, BRAIN RESEARCH, Vol. 1225,
pp. 57---66 (April 30, 2008) (hereinafter, "Lansky''). Final Act. 6-8;
Ans. 7-9.
(4) The Examiner rejected claim 10 as being unpatentable under
35 U.S.C. § 103(a) over the combination of Alvarado, Nedungadi, and
Foffani, Guglielmo et al., PSTH-based classification of sensory stimuli using
ensembles of single neurons, JOURNAL OF NEUROSCIENCE METHODS, Vol.
135, pp. 107-20 (2004) (hereinafter, "Foffani"). Final Act. 8-9; Ans. 9-10.
Appellants' Contentions
(1) Appellants contend (App. Br. 6; Reply Br. 5) the Examiner erred
in rejecting claims 1--4, 7, 11-14, 17, and 18 under 35 U.S.C. § 103(a) over
the combination of Alvarado and Nedungadi based on an interpretation that
claim 1 requires binning separate classes of neurons and not each neuron
individually. Therefore, Appellants assert that Nedungadi fails to disclose,
and as a result the base combination of Alvarado and Nedungadi fails to
teach or suggest, a method for time encoding using an IF sampler, including
"binning the pulse train to generate a feature vector for each of the separate
classes," as recited in claim 1. Specifically, Appellants contend (Reply
Br. 5) that Nedungadi discloses binning each neuron, and not separate
classes of neurons as recited in claim 1, and the portions of Appellants'
Specification cited by the Examiner in the response to Appellants'
3
Appeal2014-009124
Application 13/157,009
arguments in the Appeal Brief (see Ans. 12-13 citing Nedungadi iii! 13, 25,
29, and 36; Figs. 4a, 4b, and 5) make it clear that classes are subsets and not
every neuron, thus classes cannot be neurons.
(2) Appellants' remaining arguments as to the rejections of dependent
claims 5, 6, 10, 15, and 19 are based on the same interpretation of "classes"
as argued for claim 1.
(3) Appellants also contend (App. Br. 9-11) that Lansky fails to
(a) act based on a feature vector as recited in claim 8; (b) discuss a
likelihood ratio as recited in claim 9; and ( c) assign classes based on
conditional probability densities as recited in claims 16 and 20.
Issues on Appeal
Based on Appellants' arguments in the Appeal Brief (App. Br. 5-12)
and the Reply Brief (Reply Br. 3-7), the following two issues are presented
on appeal:
(1) Did the Examiner err in rejecting claims 1-7, 10-15, and 17-19 as
being obvious over the base combination of Alvarado and Nedungadi
because (a) a proper interpretation of representative claim 1 requires binning
separate classes of neurons and not each neuron individually; and (b) thus
Nedungadi fails to disclose, and as a result the base combination of Alvarado
and N edungadi fails to teach or suggest, a method for time encoding using
an IF sampler, including "binning the pulse train to generate a feature vector
for each of the separate classes," as recited in representative claim 1?
(2) Did the Examiner err in rejecting claims 8, 9, 16, and 20 as being
obvious because Lansky fails to teach or suggest the limitations at issue of
(i) acting based on a feature vector (claim 8); (ii) using a likelihood ratio
(claims 9 and 20); and/or (iii) assigning classes based on conditional
4
Appeal2014-009124
Application 13/157,009
probability densities (claim 16), as set forth in the respective dependent
claims?
ANALYSIS
We have reviewed the Examiner's rejections (Final Act. 3-9;
Ans. 3-10) in light of Appellants' contentions in the Appeal Brief (App.
Br. 5-12) and the Reply Brief (Reply Br. 3-7) that the Examiner has erred,
as well as the Examiner's response to Appellants' arguments in the
Examiner's Answer found at pages 11-17. We disagree with Appellants'
conclusions.
Claims 1-7, 10--15, and 17-19
With regard to representative independent claim 1, we adopt as our
own ( 1) the findings and reasons set forth by the Examiner in the action
from which this appeal is taken (Final Act. 3--4; Ans. 4) with the exception
that we agree with the Examiner (Ans. 4) the final vector D ofNedungadi is
equivalent to the recited feature vector in claim 1 (and not the vector
counting variable N(t) relied on by the Examiner at Final Act. 3), and (2) the
reasons set forth by the Examiner in the Examiner's Answer in response to
Appellants' Appeal Brief (see Ans. 11-13). Based on our interpretation of
the term "classes" recited in representative claim 1, we concur with the
conclusions reached by the Examiner with regard to the obviousness of
representative claim 1 in view of the base combination of Alvarado and
Nedungadi.
"During examination, 'claims ... are to be given their broadest
reasonable interpretation consistent with the specification, and ... claim
language should be read in light of the specification as it would be
interpreted by one of ordinary skill in the art."' In re Am. Acad. of Sci. Tech
5
Appeal2014-009124
Application I3/157,009
Ctr., 367 F.3d I359, I364 (Fed. Cir. 2004) (quoting Jn re Bond, 9IO F.2d
83I, 833 (Fed. Cir. I990)); see also In re Morris, I27 F.3d I048, I053-54
(Fed. Cir. I997).
In this light, we agree with the Examiner (Ans. I2-I3) that the claims
on appeal do not actually require or contain limitations requiring a class to
be a subset or a plurality of neurons, as opposed to individual neurons.
Claim I on appeal simply recites a method of encoding by "receiving input
signals for separate classes," using the input signals to generate a pulse
train, and "binning the pulse train to generate a feature vector for each of the
separate classes" (claim I) (emphases added). Claim I does not specify
what the classes are, or what the classes consist of, other than to state that a
feature vector is generated "for each of the separate classes" (claim I).
Turning to Appellants' Specification, we find Figures 4a and 4b show
two separate classes (Class I in Fig. 4a and Class 2 in Fig. 4b ), and Figure 5
shows a distribution of projections of two classes NI and N2 taken from
Figures 4a and 4b (see Spec. i-fi-13 I-36). The Specification is relatively
silent as to the composition of the classes (see Spec. i-fi-f I 7, 25, 26, 29, 3 I, 45
all describing the classes as being separate; Fig. 2 showing class assignment
module 220 operating as described in i-f3 I). However, paragraph I 3 of
Appellants' Specification does describe that the IF sampler uses time-based
encoding with "linear discriminant analysis" to discriminate features of the
input signal (such as "action potentials from two separate neurons"). It
follows that Appellants disclose two separate classes of neurons (e.g.,
Classes I and 2 shown in Figs. 4a and 4b and neurons NI and N2 shown in
Fig. 5) are discriminated for subsequent use in generating a feature vector
from a pulse train generated using the input signals, each having a different
6
Appeal2014-009124
Application 13/157,009
action potential from the other. Therefore, the Examiner's interpretation that
classes are neurons as supported by paragraphs 13, 25, 29, and 36 and
Figures 4a, 4b, and 5 is reasonable.
Final vector Din Nedungadi is a "binned data vector" (Nedungadi,
p. 57, Section 2.1) generated from a group of m neurons including ith andjth
neurons (Nedungadi, p. 56, Section 2.1). In view of the Specification's
descriptions (or lack thereof) of the separate classes discussed supra, we
agree with the Examiner (Ans. 12-13) that Nedungadi's ith andjth neurons
meet the claim language of "separate classes" set forth in claim 1. As a
result of this interpretation, which is supported by Appellants' Specification
and Drawings, we find Nedungadi discloses binning a pulse train to generate
a final vector for separate classes as recited in claim 1. We also agree with
the Examiner's findings regarding Alvarado (see Final Act. 3; Ans. 4), and
note the striking similarity of Appellants' Figure 1 and Alvarado's Figure 1.
The base combination of Alvarado and Nedungadi therefore teaches, or at
least suggests, the method for time encoding using an IF sampler, including
"binning the pulse train to generate a feature vector for each of the separate
classes," as recited in representative claim 1.
In view of the foregoing, we sustain the Examiner's rejection of
representative independent claim 1 over the base combination of Alvarado
and Nedungadi, as well as claims 2--4, 7, 11-14, 17, and 18 grouped
therewith. For similar reasons, and because Appellants ultimately rely on
the same arguments as to claim 1 for the patentability of dependent claims 5,
6, 10, 15, and 19 (see App. Br. 8-9 and 12), we also sustain the Examiner's
rejections of (i) claims 5, 6, 15, and 19 over the combination of Alvarado,
7
Appeal2014-009124
Application 13/157,009
Nedungadi, and Sanchez; and (ii) claim 10 over the combination of
Alvarado, Nedungadi, and Foffani.
Claims 8, 9, 16, and 20
Appellants' contentions (App. Br. 9-11) that the Examiner erred in
determining the combination of Alvarado, Nedungadi, and Lansky teaches
or suggests the methods or systems for time encoding using an IF sampler
recited in claims 8, 9, 16, and 20 are unpersuasive.
With regard to claim 8, we agree with the Examiner (Final Act. 6-7;
Ans. 16) that (i) the base combination of Alvarado and Nedungadi, and not
Lansky, is relied upon as teaching or suggesting acting based on a feature
vector; and (ii) Lansky (pages 57-58; Fig. 2, showing probability functions)
teaches or suggests using a conditional probability density to assign classes.
With regard to claim 9, we agree with the Examiner (Final Act. 7;
Ans. 8 and 16) that Lansky teaches or suggests the equivalent of the recited
likelihood ratio at page 59, section 2.2 by determining response distributions
and probability density functions as shown in Figure 2. Appellants'
contention (App. Br. 10) that Lansky fails to discuss a likelihood ratio is not
persuasive and no evidence has been presented, only attorney argument, to
show that Lansky's response distributions and probability density functions
are not the equivalent of the recited likelihood ratio of a conditional
probability density.
With regard to claims 8, 16, and 20, we agree with the Examiner
(Final Act. 6-8; Ans. 7-9 and 16-17) that the base combination of Alvarado
and Nedungadi, and not Lansky, is relied upon as teaching or suggesting
assigning classes based on conditional probability densities. In view of the
8
Appeal2014-009124
Application 13/157,009
foregoing, we sustain the Examiner's rejection of claims 8, 9, 16, and 20
over the combination of Alvarado, Nedungadi, and Lansky.
CONCLUSIONS
( 1) The Examiner has not erred in rejecting claims 1-7, 10-15, and
17-19 as being obvious over the base combination of Alvarado and
Nedungadi because (a) representative claim 1 does not require separate
classes of neurons as Appellants argue (Reply Br. 4), but instead merely
receives "input signals for separate classes" without specifying what the
classes are for or consist of other than to state that a feature vector is
generated "for each of the separate classes" (claim 1 ); and (b) N edungadi
discloses, and the base combination of Alvarado and Nedungadi teaches or
suggests, the method for time encoding using an IF sampler, including
"binning the pulse train to generate a feature vector for each of the separate
classes," as recited in representative claim 1.
(2) Appellants have not shown the Examiner erred in rejecting
claims 8, 9, 16, and 20 as being unpatentable under 35 U.S.C. § 103(a).
Accordingly, we affirm the obviousness rejections of claims 8, 9, 16, and 20
over Alvarado, Nedungadi, and Lansky.
DECISION
We affirm the Examiner's rejections of claims 1-20.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(l )(iv).
AFFIRMED
9