`FROM DR. MITZENMACHER’S DEPOSITION1 AND RELEVANCE TO DECLARATION
`Citations
`Relevance to Mitzenmacher Declaration
`
`25:12-27:24;
`28:19-30:14;
`32:12-35:24;
`39:24-42:9;
`44:24-45:10;
`51:10-52:15;
`55:16-63:6
`
`45:11-51:9
`
`64:22-73:14;
`74:19-78:18;
`
`Dr. Mitzenmacher opined that Luby’s and MacKay’s
`irregularity could be confined to parity bits, such that it would
`not lead to irregular use of information bits. See IPR2017-
`2192, Ex. 2004 ¶¶ 74-77, 84-85, 88; IPR2017-297, Ex. 2004 ¶¶
`73-78. These questions elicit testimony that Luby’s and
`MacKay’s irregularity do relate to irregular use of information
`bits. Systematic versions of both encoders were obvious and in
`systematic versions, the irregularity would relate to information
`bits.
`Dr. Mitzenmacher offered opinions regarding obviousness. See
`IPR2017-210, Ex. 2004 ¶¶ 87-125. Questions aimed at
`assessing Dr. Mitzenmacher’s level of experience with
`encoders relate to his qualifications for offering opinions on
`obviousness.
`Dr. Mitzenmacher also offered opinions regarding secondary
`considerations of non-obviousness. See id., ¶¶ 126-129. In
`particular, Dr. Mitzenmacher opined that the DVB-S2 standard
`practiced the claims. See id., ¶¶ 130-137. However, a
`communication standard such as DVB-S2 cannot practice the
`claims. Only an implementation of the standard, or standard
`compliant product, could practice the claims. Dr.
`Mitzenmacher’s testimony regarding his experience with
`encoder implementations, and his failure to consider DVB-S2
`implementations, was therefore relevant to his opinion about
`secondary considerations.
`Dr. Mitzenmacher purported to distinguish Frey with reference
`to Figure 3 of the patents. See IPR2017-210, Ex. 2004 ¶¶ 76-
`
`1 Exhibit 1062 in IPR2017-00210, which corresponds with Exhibit 1262 in IPR2017-00219,
`Exhibit 1045 in IPR2017-00297 and Exhibit 2038 in IPR2017-00700, -00701, and -00728.
`2 Dr. Mitzenmacher’s deposition was taken simultaneously for IPR2017-00210, -219, -297, -700,
`-701, and -728. Petitioner addresses these related cases herein because Caltech combines its
`arguments for all six cases in its motion.
`
`APPLE - EXHIBIT 1274
`Apple v. Caltech
`IPR2017-00728
`
`1
`
`
`
`
`
`Citations
`
`80:13-81:21;
`82:3-86:7;
`86:18-87:14;
`88:24-89:10;
`111:14-112:12
`131:12-135:5
`
`145:6-146:20;
`147:2-155:16;
`163:12-168:8;
`401:14-404:2;
`
`156:12-162:8;
`163:12-168:8
`
`181:3-192:10;
`
`202:10-207:13
`
`
`
`
`Relevance to Mitzenmacher Declaration
`
`86. These questions and Exhibit 1044 relate to showing that
`(a) the claims are broader than Figure 3 and (b) Dr.
`Mitzenmacher had an overly narrow view of the scope of the
`claims.
`
`Dr. Mitzenmacher purports to distinguish the prior art based on
`Figure 3 of the patents. See IPR2017-210, Ex. 2004 ¶¶ 76-86.
`These questions relate to establishing claim scope and showing
`that the claims are broader than Figure 3.
`Dr. Mitzenmacher opined that modifying Divsalar to be
`irregular would have been difficult and a POSA would not
`have had an expectation of success. See IPR2017-219, Ex.
`2004 ¶¶ 104-107, 115-125. These questions relate to the
`simplicity of modifying Divsalar to make it irregular and a
`POSA’s expectation of success.
`Dr. Mitzenmacher opined on the Khandekar thesis. IPR2017-
`219, Ex. 2004 ¶¶ 108-114. These questions relate to that
`opinion.
`Dr. Mitzenmacher opined that modifying Divsalar in view of
`Luby would involve changing Divsalar’s accumulator and that
`Divsalar cannot be made random without changing the
`accumulator. See IPR2017-219, Ex. 2004 ¶¶ 86-87, 112, 124.
`Also, like Divsalar, Ping has an accumulator. Dr.
`Mitzenmacher opined that, because of this accumulator, Ping is
`already irregular and a POSA would therefore have not used
`MacKay’s irregularity in Ping. See IPR2017-297, Ex. 2004 ¶¶
`81-92. These questions about the accumulator relate to
`rebutting those opinions, showing the simplicity of the
`accumulator and that a POSA would have modified other parts
`of Divsalar’s and Ping’s code instead of the accumulator.
`Dr. Mitzenmacher opined that modifying Divsalar in view of
`Luby would involve changing Divsalar’s accumulator. See
`IPR2017-219, Ex. 2004 ¶¶ 86-87, 112, 124. These questions
`elicit testimony that undermines Dr. Mitzenmacher’s opinion
`by showing that a POSA would have modified Divsalar’s
`repeater instead.
`
`2
`
`
`
`
`
`Citations
`
`228:21-233:22
`
`259:5-265:13;
`267:8-273:2;
`276:21-278:2;
`278:9-282:21;
`283:14-21;
`284:24 (the
`phrase “and
`193Y”)
`404:16-407:2
`
`413:24-418:13
`
`418:14-424:14
`
`
`
`
`Relevance to Mitzenmacher Declaration
`
`Dr. Mitzenmacher opined that Luby’s teachings regarding
`irregularity could be confined to parity bits, such that a POSA
`would not have understood that irregularity could be applied to
`information bits. See IPR2017-219, Ex. 2004 ¶¶ 74-77 (“[A]
`person of ordinary skill in the art would not have thought our
`‘irregular bipartite graphs’ teaches anything about how the
`information bits are to be treated during generation of the
`codeword, whether through repetition or not.”), 84-85, 88.
`These questions relate to rebutting that opinion by showing that
`a POSA would have been motivated to make Luby’s
`information bits irregular, not just the parity bits.
`Dr. Mitzenmacher opined that MacKay’s irregularity could be
`confined to parity bits, such that it would not lead to irregular
`use of information bits. See IPR2017-297, Ex. 2004 ¶¶ 73-78.
`These questions relate to rebutting that opinion by showing that
`MacKay does teach irregular use of information bits.
`
`Dr. Mitzenmacher purports to distinguish Frey based on the
`code rate, arguing that increasing repetition in Frey increases
`the size of the codeword. See IPR2017-210, Ex. 2004 ¶¶ 82-
`83. These questions relate to rebutting that argument by
`showing the patent discloses and claims encoders that would
`also increase the codeword in response to increasing repetition.
`Dr. Mitzenmacher opined that a POSA would not have
`combined Divsalar and Luby. IPR2017-219, Section VIII. He
`also opined that Divsalar and Luby are different types of codes.
`Id. at ¶¶ 97, 101. These questions and Exs. 1046-1047 relate to
`rebutting that opinion by showing the similarity of Divsalar’s
`and Luby’s code.
`Dr. Mitzenmacher opined that a POSA would not have
`combined Divsalar and Luby. IPR2017-219, Section VIII. He
`also opined that Divsalar and Luby are different types of codes.
`Id. at ¶¶ 97, 101. These questions and Exs. 1046-1047 relate to
`
`3
`
`
`
`
`
`Citations
`
`Relevance to Mitzenmacher Declaration
`
`rebutting that opinion by showing the similarity of Divsalar’s
`and Luby’s code.
`Dr. Mitzenmacher opined that a POSA would not have
`combined Ping and MacKay. IPR2017-297, Section IX(C).
`These questions and Exs. 1048-1049 relate to rebutting that
`opinion by showing the similarity of Ping’s and MacKay’s
`code.
`Dr. Mitzenmacher opined that a POSA would not have
`combined Ping and MacKay. IPR2017-297, Section IX(C).
`These questions and Exs. 1048-1049 relate to rebutting that
`opinion by showing the similarity of Ping’s and MacKay’s
`code.
`Dr. Mitzenmacher offers opinions regarding WiFi standards.
`See IPR2017-00210, Ex. 2004 ¶ 120. In particular, Dr.
`Mitzenmacher opined that “Divsalar is already too slow for
`many practical applications, such as 802.11.” These questions
`relate to Dr. Mitzenmacher’s familiarity with the standard and
`its implementation.
`
`
`424:15-431:24
`
`431:25-438:24
`
`445:11-446:12
`
`
`
`ALLEGED OUT-OF-SCOPE QUESTIONS
`FROM DR. DIVSALAR’S DEPOSITION3 AND RELEVANCE TO DECLARATION
`Citations
`Relevance to Dr. Divsalar Declaration
`
`23:7-25:24
`
`Dr. Divsalar opined that research on irregular LDPC codes was
`concerned with modifying traditional Gallager codes and a
`POSA would have no motivation to apply such a teaching to
`RA codes. See IPR2017-002104 & IPR2017-00219, Ex. 2031 ¶
`10. Additionally, Dr. Divsalar’s declaration attaches his CV,
`
`3 Ex 1064 of IPR2017-00210, which corresponds with Exhibit 1264 in IPR2017-00219 and
`Exhibit 2039 in IPR2017-00700, -00701, and -00728.
`4 Dr. Divsalar’s deposition was taken simultaneously for IPR2017-00210, -219, -700, -701, and -
`728. Petitioner addresses these related cases herein because Caltech combines its arguments for
`all six cases in its motion.
`
`
`
`
`4
`
`
`
`
`
`Citations
`
`Relevance to Dr. Divsalar Declaration
`
`which contends that he “[m]ade significant contribution to
`channel coding for deep space communications [and p]roposed
`new turbo codes and protograph-based LDPC codes to be used
`in future NASA missions. The proposed codes became
`CCSDS standard for space applications.” Ex. 2032 at 3. These
`questions relate to Dr. Divsalar’s background and knowledge
`of the CCSDS standard, on which he worked, and the focus of
`the LDPC codes research at that time in order to establish
`whether a POSA would have a motivation to modify RA codes.
`
`Dr. Divsalar discussed submitting a paper entitled, “Coding
`Theorems for ‘Turbo-Like’ Codes,” in connection with the
`Allerton conference in 1998. See IPR2017-00210 & IPR2017-
`00219, Ex. 2031 ¶ 19. These questions relate to what “in
`connection with the Allerton conference” means.
`Dr. Divsalar offered opinions regarding potential combinations
`and modifications of RA codes based on his paper, “Coding
`Theorems for ‘Turbo-Like’ Codes.” See IPR2017-00210 &
`IPR2017-00219, Ex. 2031 ¶¶ 1, 27, 33-35. These questions
`relate to how data is partitioned based on Figure 3 in Dr.
`Divsalar’s paper, which displays an encoder for a repeat-
`accumulate code, and what a POSA would understand from
`interpreting this information.
`
`Dr. Divsalar opined that it would not have been trivial or
`obvious to modify RA codes by making them “irregular” in
`order to arrive at IRA codes and that a POSA would not be
`motivated to make such a modification. See IPR2017-00210 &
`IPR2017-00219, Ex. 2031 ¶¶ 9 (“I do not believe it would have
`been trivial or obvious … nor would a person of ordinary skill
`in the art be motivated to make such a modification.”), 33-35.
`These questions relate to refuting Dr. Divsalar’s opinion on
`whether it would have been trivial or obvious to modify RA
`codes to make them irregular.
`Dr. Divsalar testified that he was “aware the Tanner graphs
`could be used to represent LDPC codes [but] we did not
`consider Tanner graph representation useful or applicable to
`concatenated convolutional codes.” See IPR2017-00210 &
`
`29:1-36:18
`
`48:20-49:19,
`50:13-54-20
`
`56:12-58:2,
`58:22-68:9, 69:8-
`18
`
`77:23-78:23,
`82:5-93:16
`
`
`
`
`5
`
`
`
`
`
`Citations
`
`Relevance to Dr. Divsalar Declaration
`
`IPR2017-00219, Ex. 2031 ¶ 26. These questions and Ex. 1057
`relate to establishing whether a POSA would have considered
`Tanner graphs useful and applicable for representing
`concatenated convolutional codes at that time.
`Dr. Divsalar offered opinions regarding obviousness and
`whether a POSA would apply irregular repetition of
`information bits to an RA code. See IPR2017-00210 &
`IPR2017-00219, Ex. 2031 ¶ 36. Specifically, Dr. Divsalar
`expressed that “applying irregular repetition of information bits
`to an RA code would actually be expected to degrade
`performance of the code” because it would “drastically reduce
`the coding rate.” Id. These questions and Exs. 1058 and 1057
`rebut this opinion by demonstrating that a POSA would have
`known how to modify a regular code to perform irregular
`repetition without changing the coding rate.
`Dr. Divsalar opined that “[m]aking the RA codes systematic
`would be expected to decrease the code rate” (IPR2017-00210
`& IPR2017-00219, Ex. 2031 ¶ 28), that RA codes “would have
`been more complicated by including a systematic code” (id. at
`¶ 29), and that “[t]here was no point in making RA code
`systematic.” (id. at ¶ 11). These questions relate to challenging
`Dr. Divsalar’s knowledge and opinions about systematic codes,
`including whether RA codes would be “more complicated” and
`whether “there was no point” to making codes systematic.
`Dr. Divsalar opined that “RA codes were designed to be
`research tools” and that it “would have been more complicated
`by including a systematic code.” See IPR2017-00210 &
`IPR2017-00219, Ex. 2031 ¶¶ 29-30. These questions and Ex.
`1059 relate to rebutting this opinion, and showing how a POSA
`could modify the RA code to produce a systematic codeword
`without complication. See, e.g., 116:21-117:3, 120:2-14.
`Dr. Divsalar opined that “it would not have been obvious to
`modify my RA codes to include an irregular repeat” and that
`“there would be no motivation to do so” because “[t]he
`contemporaneous technical literature on ‘irregular codes’ was
`directed to specific types of irregularity that had no meaningful
`
`94:9-101:22,
`102:17-107:13
`
`107:18-110:2,
`112:23-114:5
`
`114:6-120:17
`
`128:12-130:19,
`131:18-136:12,
`141:10-23,
`142:22-143:8,
`145:4-156:1,
`
`
`
`
`6
`
`
`
`
`
`Citations
`
`161:15-167:9,
`188:5-208:5,
`235:11-237:25,
`239:3-252:3,
`252:14-22,
`254:4-257:13,
`271:12-274:24
`
`143:22-145:3
`
`156:2-161:13
`
`168:5-178:7
`
`
`
`
`Relevance to Dr. Divsalar Declaration
`
`application to RA codes.” IPR2017-00210 & IPR2017-00219,
`Ex. 2031 ¶¶ 9-10; see also id., ¶¶ 33-35. These questions relate
`to what prior art Dr. Divsalar considered in forming this
`opinion, whether this opinion fairly presents the state of
`“contemporaneous technical literature,” and whether a POSA
`would find that disclosure of irregular codes in Luby98 and
`Frey, which Dr. Divsalar conceded to be contemporaneous
`technical literature, could be meaningfully applied to RA code.
`Dr. Divsalar offered opinions regarding obviousness and how it
`would be a “difficult and complex task to determine how to
`apply irregularity [to RA codes] in a manner that is effective.”
`See IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 34. These
`questions relate to whether it would have been difficult or
`complex for a POSA to apply irregularity to RA codes.
`Dr. Divsalar opined that “applying irregular repetition of
`information bits to an RA code would actually be expected to
`degrade performance of the code” by reducing the “coding
`rate,” and this would have deterred a POSA from applying
`irregular repetition of information bits to RA codes. See
`IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶¶ 35, 36.
`These questions and Ex. 1057, which is a Tanner graph
`representation of an RA code where Q equals 5, and Ex. 1058,
`where Q equals 3 for half the information nodes and Q equals 7
`for the other half, relate to rebutting Dr. Divsalar’s opinion.
`Specifically, these questions and exhibits establish that regular
`RA code can be modified to IRA code without changing the
`code rate (and that there are a multitude of ways to do so).
`Dr. Divsalar offered opinions regarding message passing
`decoding as discussed in his paper, “Serial Concatenation of
`Interleaved Codes: Performance Analysis, Design, and
`Iterative Decoding.” See IPR2017-00210 & IPR2017-00219,
`Ex. 2031 ¶¶ 3, 32. These questions relate to what a POSA
`would know about “message-passing decoders” as of the time
`of Dr. Divsalar’s paper, and thus what such a POSA would
`have understood from reading this teaching in that paper.
`
`7
`
`
`
`
`
`Citations
`
`180:14-181:15
`
`181:16-187:16
`
`208:6-213:10,
`214:22-222:21,
`223:21-225:24
`
`225:25-231:10,
`232:12-234:1
`
`
`
`
`Relevance to Dr. Divsalar Declaration
`
`Dr. Divsalar asserted that his paper did not discuss “repeating
`some bits a different number of times from others, nor did we
`contemplate such an arrangement at the time we wrote the
`paper.” See IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 27.
`These questions relate to whether modifying RA codes to
`repeat some bits a different number of times from others (i.e.,
`irregularly) was in fact suggested to Dr. Divsalar by Dr. Frey
`(or others) in the prior art period, and whether Dr. Divsalar
`contemplated Dr. Frey’s teaching at that time.
`Dr. Divsalar asserted that his paper did not discuss “repeating
`some bits a different number of times from others, nor did we
`contemplate such an arrangement at the time we wrote the
`paper.” See IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 27.
`These questions relate to whether modifying RA codes to
`repeat some bits a different number of times from others (i.e.,
`irregularly) was in fact suggested to Dr. Divsalar by Dr. Frey
`(or others) in the prior art period, and whether Dr. Divsalar
`contemplated Dr. Frey’s teaching at that time.
`Dr. Divsalar opined that “[i]t would not have been trivial or
`obvious to a person of ordinary skill to modify my RA code
`paper to create IRA codes” and that it would be a “difficult and
`complex task to determine how to apply irregularity in a
`manner that is effective.” See IPR2017-00210 & IPR2017-
`00219, Ex. 2031 ¶¶ 33, 34. These questions and Ex. 1060
`relate to rebutting Dr. Divsalar’s obviousness opinion by
`showing that it would have been straightforward to modify the
`repeater of Dr. Divsalar’s Fig. 3 so as to incorporate the
`irregular repetition taught by Dr. Frey’s Fig. 1, and that doing
`so would have been effective.
`Dr. Divsalar stated that he did “not believe it would have been
`trivial or obvious to modify RA codes by making them
`‘irregular’ in order to arrive at IRA codes” and that a POSA
`would not have been motivated to make such a modification.
`See IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 9. These
`questions relate to whether it would have been obvious to a
`POSA to modify RA codes by making them irregular.
`
`8
`
`
`
`
`
`Citations
`
`258:4-260:12
`
`260:13-268:15
`
`Relevance to Dr. Divsalar Declaration
`
`Dr. Divsalar noted that “RA codes are not a subset or
`simplification of turbo codes, as they rely on serial
`concatenation rather than parallel concatenation.” See
`IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 24. These
`question and Ex. 1061 relate to rebutting Dr. Divsalar’s
`opinion by showing that IRA codes generalize RA codes by
`allowing irregular repetition of the input bits. This cross-
`examination thus impeaches Dr. Divsalar’s opinion with his
`own prior inconsistent statements.
`Dr. Divsalar opined that “[p]eople from the coding community
`in the late 1990s generally did not think of concatenated
`convolutional codes in terms of their parity check matrices.”
`See IPR2017-00210 & IPR2017-00219, Ex. 2031 ¶ 26. These
`questions relate to rebutting Dr. Divsalar’s opinion by showing
`that POSAs in 1999 knew how to represent codes in terms of
`parity check matrices and generator matrices, that any linear
`block code can be represented by a generator matrix or parity
`check matrix, and that a POSA would have known in 1999 that
`the first coder in Dr. Divsalar’s Fig. 3 can be represented as a
`low-density generator matrix.
`
`
`
`
`
`
`9
`
`