`2
`3
`4
`5
`6
`7
`8
`9
`
`10
`
`11
`
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`22
`23
`24
`25
`
`Page 1
`
`Michael Mitzenmacher, PhD
` IN THE UNITED STATES PATENT AND TRADEMARK
`OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD
`APPLE INC.,
`PETITIONER
`v.
`CALIFORNIA INSTITUTE OF TECHNOLOGY
`PATENT OWNER
`
`IPR2017-00210
`IPR2017-00700
`IPR2017-00219
`IPR2017-00728
`IPR2017-00701
`IPR2017-00297
`
`DEPOSITION OF
`MICHAEL MITZENMACHER, PhD, Volume I
`Portland, Oregon
`Wednesday, December 20, 2017
`
`Reported by:
`LISA TRONCOSO, RPR, CSR, CLR
`JOB NO. 135227
`
`TSG Reporting - Worldwide 877-702-9580
`
`CALTECH - EXHIBIT 2038
`Apple Inc. v. California Institute of Technology
`IPR2017-00728
`
`
`
`Michael Mitzenmacher, PhD
`December 20, 2017
`9:00 a.m.
`
`Page 2
`
`Deposition of MICHAEL MITZENMACHER,
`PhD, Volume I, held at the offices of Wilson,
`Sonsini, Goodrich & Rosati, 701 Fifth Avenue,
`Suite 5100, Seattle, Washington, before
`Lisa Troncoso, a Registered Professional
`Reporter, Certified Livenote Reporter, and
`Oregon Certified Shorthand Reporter
`Number 14-0429.
`
`TSG Reporting - Worldwide 877-702-9580
`
`1
`2
`3
`
`4 5
`
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
`
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
`Page 3
`
` Michael Mitzenmacher, PhD
`A P P E A R A N C E S
`WILMERHALE
`Attorneys for Petitioner
` 60 State Street
` Boston, MA 02109
`BY: RICHARD GOLDENBERG, ESQ.
` - AND -
` MICHAEL SMITH, ESQ.
`
`Wilson Sonsini Goodrich & Rosati
`Attorneys for Patent Owner
` 701 Fifth Avenue
` Seattle, Washington 98104
`BY: MICHAEL ROSATO, ESQ.
` - AND -
` QUINCY LU, ESQ.
` - AND -
` MATTHEW ARGENTI, ESQ. (via telephone)
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 4
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` -O0O-
` MICHAEL MITZENMACHER, PhD,
`called as a witness, having first been sworn by
`the Certified Shorthand Reporter, was examined
`and testified as follows:
` MR. ROSATO: Before you start, just
` so you -- I'm just organizing stuff and
` noticed this. I think these are just --
` looks like just clean copies of his
` declaration, if you want to take a look.
` MR. GOLDENBERG: Let me just take a
` peak. That's fine, that will spare us some
` exhibits.
` (Michael Mitzenmacher's folder of
` Declarations, marked.)
` MR. GOLDENBERG: So, sorry, we did
` that on the record, right?
` MR. ROSATO: It looks like we did
` actually.
`EXAMINATION BY:
`MR. GOLDENBERG:
` Q. Dr. Mitzenmacher, good morning.
` A. Good morning.
` Q. My name is Richard Goldenberg and
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 5
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`I'll be taking your deposition today.
` A. Okay.
` Q. Have you been deposed before?
` A. Yes.
` Q. How many times?
` A. Many. Like, I don't recall, but
`probably more than twenty.
` Q. All for patent cases?
` A. Some of the cases involved like
`trade secrets or copyright, most of them have
`been patent cases.
` Q. And where are you presently
`employed?
` A. Harvard University.
` Q. And are you doing a post in Seattle?
`I'm just curious why the deposition is in
`Seattle and not Boston.
` A. The lawyer sort of said, you know,
`where would you like to do it and I really
`didn't have a preference. You know, my parents
`lived in L.A. on the west coast, so I said I'd
`be more than happy to come out on the west
`coast given the cold weather in Boston, and
`then give me a chance -- I stopped and saw my
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 6
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`parents before coming up here which, you know,
`was a nice opportunity for me.
` Q. Sounds good. I was just curious.
` So I was -- just before I saw the
`binder that your counsel showed me, I was going
`to give you one of the -- one of your
`declarations, but you already have a copy.
` A. However you want to do it. Whatever
`is easier for you.
` Q. Can you turn to the declaration for
`the '781 patent in your binder there?
` MR. ROSATO: Counsel, if you have
` copies, can I also have a copy?
` MR. GOLDENBERG: Sure.
` MR. ROSATO: Thank you.
` THE WITNESS: The '781, so for the
` IPR00297?
`BY MR. GOLDENBERG:
` Q. Correct.
` A. I got it.
` Q. And if we just turn to the last
`page, page 88, you see it's signed Michael
`Mitzenmacher, PhD?
` A. Uh-huh.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 7
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. Is that your signature?
` A. Yes.
` Q. And you authored this declaration?
` A. Yes.
` Q. And you authored the other
`declarations that are at issue in this case?
` A. Yes.
` MR. ROSATO: Objection, form.
`BY MR. GOLDENBERG:
` Q. Maybe we can just go through your
`binder. Did you author the declaration of
`Michael Mitzenmacher for IPR2017-00210 that's
`labeled as Exhibit 2004 in that IPR?
` A. Sorry, which one again?
` Q. You probably want to look at the
`first page.
` A. 2017-210. Yes, that's my signature,
`and this is my report expressing my opinions
`that I wrote and worked on with counsel.
` MR. ROSATO: Counsel, I don't want
` to cause unnecessary time, just noting that
` there are other declarations besides
` Mitzenmacher declarations.
` MR. GOLDENBERG: Fair enough.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 8
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` MR. ROSATO: That's all.
` MR. GOLDENBERG: Yup.
`BY MR. GOLDENBERG:
` Q. Just to capture this in one quick
`question for the record, Dr. Mitzenmacher, you
`were the author of declaration of Michael
`Mitzenmacher for IPR2017-00210 that's marked as
`Exhibit 2004, correct?
` A. Yes. Reporting my opinions, I
`authored it and obviously I worked on it with
`counsel.
` Q. And just flip to the declaration for
`the next IPR2017-219.
` A. Yes.
` Q. So you were the author of the
`declaration of Michael Mitzenmacher for
`IPR2017-00219 that's marked as Exhibit 2004,
`correct?
` A. Yes.
` Q. And if you turn to the next
`declaration that's the IPR2017-00297.
` A. Yes.
` Q. So, Dr. Mitzenmacher, you're also
`the author of the declaration of Michael
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 9
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`Mitzenmacher for IPR2017-00219 that's marked as
`Exhibit 2004, correct?
` A. Yes.
` Q. And if you turn to -- let me start
`again. Are you ready?
` A. Yeah.
` Q. Do you have your declaration for
`IPR2017-700, 701 and 728?
` A. Yes.
` Q. So you are the author of the
`declaration of Dr. Michael Mitzenmacher for
`IPR2017-700, 2017-701 and 2017-728 that's
`marked as Exhibit 2004, correct?
` A. Yes. So same answer, I am the
`author, these express my opinions and, again,
`obviously I worked on them with counsel in
`preparing the reports.
` Q. Understood. Thank you.
`Dr. Mitzenmacher, every parity-check matrix has
`a corresponding tanner graph, correct?
` MR. ROSATO: Objection, form.
` THE WITNESS: I'd say that you can
` take a parity-check matrix and one of the
` ways you can express it is through a tanner
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 10
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` graph. As a scientist or engineer I always
` hesitate to say every, but scientists don't
` always like to work in absolutes, but I
` can't think of an example of a, you know,
` certainly standard parity-check matrix that
` you cannot express as a tanner graph.
`BY MR. GOLDENBERG:
` Q. Dr. Mitzenmacher, just to make sure
`I understood your testimony there, if one has a
`parity-check matrix one can generate a tanner
`graph that expresses the same code, correct?
` A. If one has a parity-check matrix,
`again, in the manner described, for example, in
`some of my work, and if you're using the same
`definition of a parity-check matrix then, yes,
`one of the ways you can express the code that
`that matrix represents is also through a tanner
`graph.
` Q. And that works in the reverse, too,
`correct. If you have a tanner graph you could
`generate a corresponding parity-check matrix,
`correct?
` A. I would say only for limited classes
`of tanner graphs that would depend on how you
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 11
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`were using the term. Typically, in the modern
`parlance, tanner graphs would be understood
`that you could have more complicated operations
`and simple parity checks, in which case you
`might not be able to represent that more
`complex type of operation within a parity-check
`matrix.
` Q. Now, not that previous question, but
`the one before you'd mention if we're using the
`same definition of parity-check matrix.
` A. Yes.
` Q. What is your definition of
`parity-check matrix?
` A. So I would look back, for instance,
`in some of my papers, or I can look for the
`report, but typically a parity-check matrix you
`have a matrix where the entries in the matrix
`represent that the values represent that
`certain -- what are sometimes referred to as,
`in this setting, information bits, or bits that
`are going to, or depending on the setting
`they're also referred to as message bits or
`message nodes.
` Typically I use like message nodes.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 12
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`In this context, you know, they do have message
`nodes, and you have certain constraints on the
`message nodes and parity-check matrix
`represents these constraints, the constraints
`being of the form that the sum of the values
`represented by the one entries in the row of a
`matrix, zero, generally using arithmetic, but
`that depends on the setting as well.
` Q. With that understanding of the
`parity-check matrix, is it correct to say that
`a parity-check matrix will have a corresponding
`tanner graph?
` A. Yes, in the manner that we've
`described.
` Q. Is it also correct that if a
`parity-check matrix is multiplied by a valid
`code word that the result will be zero?
` A. Yes, that's what the constraints of
`the parity-check matrix represent.
` Q. Is it also correct that if a
`parity-check matrix is multiplied by a
`generator matrix for the same code that the
`result will be zero?
` MR. ROSATO: Objection, foundation.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 13
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` THE WITNESS: Do you mean a zero
` matrix, like it won't actually be zero or
` zero valued, so you mean like a zero
` matrix?
`BY MR. GOLDENBERG:
` Q. Yes, a matrix of all zeros.
` A. Let's see, I would have to double
`check, but I think that's right.
` Q. So is it correct to say that a
`parity-check matrix will have a corresponding
`set of generator matrices?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: I don't think I would
` put it that way. I would say that there
` are relationships between parity matrices
` and generator matrices that, you know, a
` parity-check matrix in that manner
` corresponds to a code, and the generator
` matrix corresponds to a code, and there are
` various ways and means that you might
` transform one into the other that are known
` in the art.
`BY MR. GOLDENBERG:
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 14
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. And what is a generator matrix?
` A. In this context, a generator matrix
`is often used to describe a way of mapping a
`message or an input into a code word or an
`output.
` Q. Is it correct to say that a
`generator matrix multiplied by a vector of
`information bits will equal a code word for the
`code?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: I would be careful of
` using the term information bits. You know,
` the information bits, the way I understand
` them, Dr. Davis describes information bits
` as a message that's meant to be sent, and
` there are various transformations that you
` can put on these information bits before
` undergoing this sort of encoding.
` So I might prefer the terminology
` of, you know, message bits, or depending on
` the context, again, assuming that we're
` talking about a binary code and working on
` bits, they're codes that work on,
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 15
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` obviously, data units that are larger than
` bits.
`BY MR. GOLDENBERG:
` Q. And if this helps avoid confusion, I
`think we can restrict all questions and answers
`today to binary codes.
` A. All right.
` Q. Are you okay with that?
` A. That would be fine.
` Q. If for some reason we need to refer
`to a nonbinary code, then we'll make explicit
`reference to it, but for now we can assume that
`all questions and answers pertain only to
`binary codes, okay?
` A. Okay.
` Q. Now, I think you said -- withdrawn.
` A message bit could be either an
`information bit or a parity bit, correct?
` A. That would depend on the context or
`setting, or how people -- yes, but typically
`often it's used in that regard.
` Q. I mean, is that how you use the
`term?
` A. I believe in at least some papers
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 16
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`that's how I've used the term.
` Q. And you would not multiply a
`generator matrix times parity bits in order to
`encode a code word, correct?
` A. No, you would -- the parity bits
`would be the -- produced by the coding
`operations.
` Q. So you expressed some reticence
`about using the term information bits. What's
`a good term to describe the bits of a code word
`that in a systematic code that are not parity
`bits?
` MR. ROSATO: Objection, form.
` THE WITNESS: Sorry, if you parse a
` bunch of things, bits of the code in a
` systematic code?
`BY MR. GOLDENBERG:
` Q. Right.
` A. I guess more talking about the
`generator matrices in particular, what I'm
`talking about is independent of whether they
`were systematic or nonsystematic. You know, I
`think generally these might be referred to as
`the systematic bits.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 17
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. Okay. So is it accurate to say then
`that a generator matrix times a vector of
`systematic bits will equal a code word?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: In the context of
` systematic codes, right, where you were
` also transmitting it, that would be one way
` you can refer to it.
`BY MR. GOLDENBERG:
` Q. Well, even in a nonsystematic code
`there's some set of information that's desired
`to be transmitted, correct?
` A. Yes.
` Q. And what's a good term for that
`information?
` A. So, again, I don't mind information
`bits as representing in some sense the
`information that you end up desiring to be
`transmitted. It's just that that might not
`actually be the information that you input or
`multiply within your parlance to create the
`code word.
` There may be other transformations
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 18
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`you put on those bits beforehand before
`actually doing the matrix multiplication for
`the code, so that's something that I think was
`unclear in Dr. Davis' description.
` Q. Is it correct that a generator
`matrix times a vector of bits will equal a code
`word?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: Yes.
`BY MR. GOLDENBERG:
` Q. And what's a good term to describe
`that vector of bits?
` MR. ROSATO: Same objection.
` THE WITNESS: Again, I would call it
` perhaps just the input to the code.
`BY MR. GOLDENBERG:
` Q. So a generator matrix times a vector
`of code input equals a code word, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: I think that's good
` terminology.
`BY MR. GOLDENBERG:
` Q. And a parity-check matrix times a
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 19
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`code word will equal zero, correct?
` A. Yes, in standard codes.
` Q. And I'm looking for a way to
`describe corresponding parity check and
`generator matrices. Would it be correct to say
`that if a generator matrix times a code input
`equals a zero, and a parity-check matrix times
`all the code words that can be produced by that
`generator also equal zero, then the generator
`matrix and parity-check matrix are for the same
`code?
` MR. ROSATO: Objection, form.
` THE WITNESS: Sorry, I missed at
` least part of it. So the generator matrix
` times --
`BY MR. GOLDENBERG:
` Q. Let me withdraw the question and
`start over. Rather than me stating the answer,
`I'll ask you to state the answer.
` What's the right criteria for a
`parity-check matrix and generator matrix to be
`said to be used for the same code?
` MR. ROSATO: Objection, form,
` foundation.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 20
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` THE WITNESS: I would say typically,
` you know, if a generator matrix based code
` is using as an input some K bits, and
` outputting N bits, then the collection of K
` bit inputs will produce a collection of N
` bit outputs, the code words, and if for all
` of those code words the parity-check matrix
` times that code word leads to a zero, or
` that is that those code words and only
` those code words satisfy the constraints,
` then they represent the same code.
`BY MR. GOLDENBERG:
` Q. And every parity-check matrix will
`have a set of generator matrices that represent
`the same code as the parity-check matrix,
`correct?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: Right. If you're
` getting that they're -- conceivably can be
` multiple generator matrices that can lead
` to the same code there would be multiple
` parity check matrices that lead to the same
` R=1 code.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 21
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`BY MR. GOLDENBERG:
` Q. So multiple parity check matrices
`can define the same code, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: Yes. Again, when we
` talk about the same code, maybe we should
` be a bit clearer, you know, exactly what it
` means subject to, you know.
` You can either mean they must come
` out with the exact same code words or you
` might mean the same code words up to some
` sort of permutations of the order of the
` bits, so on.
` But I'd say under those criteria
` it's possible that multiple parity check
` matrices can define the same code certainly
` up to at least permutation of the bits.
`BY MR. GOLDENBERG:
` Q. Any LDPC code could be described by
`multiple parity check matrices, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: With that
` understanding that I stated, yes.
`BY MR. GOLDENBERG:
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. And there are also multiple
`generator matrices that could be used to
`produce code words for the same code, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: Again, with the
` understanding of, you know, we're talking
` about up to permutation of ordering things
` like that, then certainly.
` MR. GOLDENBERG: Mr. Rosato, what's
` the basis for your objection?
` MR. ROSATO: Form and foundation.
` MR. GOLDENBERG: What specifically
` about the form did you feel was incorrect?
` MR. ROSATO: And foundation. I
` mean, what code words are we talking about?
` There's a lot of generalities here. I'm
` preserving objections because there is open
` questions in terms of scope, how this
` relates to his direct testimony. If this
` ends up going somewhere then we'll deal
` with the objections accordingly.
` MR. GOLDENBERG: What specifically
` about the form of the question did you feel
` was improper?
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 23
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` MR. ROSATO: I'll have to look at
`the specific question you're asking, as you
`seem to ask a general question about all of
`the objections.
` The question is, like I said,
`there's -- the context is fairly unclear.
`I don't want to coach the witness, so we're
`very limited in the practice and limited to
`very specific types of objections.
` MR. GOLDENBERG: I just, to the
`extent I can, I want to ask questions that
`you find that's non-objectionable and
`that's why I'm asking you.
` MR. ROSATO: I understand.
` MR. GOLDENBERG: But what
`specifically about the form did you find
`improper? If there's something I can
`correct, I'll correct it.
` MR. ROSATO: There are aspects of
`the questions that are vague.
` MR. GOLDENBERG: I'm asking you
`about the specific last question. You made
`a form objection, what was your basis?
` MR. ROSATO: There are aspects that
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 24
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` are vague, where the relevance to the
` direct testimony is at a minimum, unclear,
` and this line of questioning has gone on
` for a fair amount of time without any
` specific reference to any specific document
` that's been entered into the record, relied
` upon, or discussed by Mr. Mitzenmacher in
` his direct testimony.
` MR. GOLDENBERG: Thank you, counsel.
`BY MR. GOLDENBERG:
` Q. Is it common in error correcting
`code literature to describe a code by a
`parity-check matrix?
` A. Common? I mean, I would say that's
`one of the ways that's used. I have seen it in
`some standard textbooks and so on. I mean, I
`can't speak to the wide commonality, but I
`believe that most people in coding would
`understand the parity-check matrix in that
`context.
` Q. But you're an expert in error
`correcting codes, right?
` A. Yes.
` Q. Within your expertise, have you seen
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 25
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
`many references that describe a code in terms
`of this parity-check matrix?
` MR. ROSATO: Objection, form, scope.
` THE WITNESS: Like I've said, I've
` seen in textbooks and numerous papers and
` so on, you know, so in that sense
` certainly, yes. To the extent that's
` commonality, I don't think I'd be prepared
` to judge or say, but I answered.
`BY MR. GOLDENBERG:
` Q. People of ordinary skill in error
`correcting can determine a generator matrix for
`a parity-check matrix, correct?
` MR. ROSATO: Objection, form.
` THE WITNESS: Again, I think one
` skilled in the art would be able to find
` methods for doing that. For example,
` various textbooks and so on, they did not
` know it of themselves.
`BY MR. GOLDENBERG:
` Q. And that is one of ordinary skill
`could produce a generator matrix that is for
`the same code as a parity-check matrix,
`correct?
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 26
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` MR. ROSATO: Objection, form, scope.
` THE WITNESS: I'd say generally I
` think that's possible. Although, again,
` we'd have to look at specific instances to
` see if there were any complications. I'd
` say generally I would think that would be
` something one of skill in the art could do.
`BY MR. GOLDENBERG:
` Q. And so this may seem repetitive, but
`I just want to ask the question a slightly
`different way. Are you ready?
` A. Okay.
` Q. Given a parity-check matrix, one of
`ordinary skill can find a generator matrix for
`the same code, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: Again, so given a
` specific parity-check matrix one would, in
` the skill of the art, would know methods
` that would allow one to transform it into a
` generator.
` Sorry. To use that to determine a
` corresponding generator matrix by various
` means. You know, again, whether they
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 27
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` specifically are be able to do it might
` matter on things such as the actual size
` and so on, but for methodology for
` connecting parity-check matrix and
` generator matrices, no.
`BY MR. GOLDENBERG:
` Q. And in particular, given a
`parity-check matrix for an LDPC code, one of
`ordinary skill could determine a generator
`matrix for that code, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: With the understanding
` and caveats I've described, yes, if one is
` given a specific matrix there are known
` methodologies.
`BY MR. GOLDENBERG:
` Q. In LDPC code -- withdrawn.
` For any given LDPC code there will
`always exist a generator corresponding to a
`systematic version of the code, correct?
` MR. ROSATO: Objection, form.
` THE WITNESS: I'm not clear on what
` you mean exactly by that question.
`BY MR. GOLDENBERG:
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 28
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. Well, just for clarity, what's a
`systematic code?
` A. A systematic code is one in which
`we're referring to before, I guess, as to the
`code bits are transmitted as part of the code
`word or form part of the code word.
` Q. So in a systematic code, the input
`of the code forms part of the code word,
`correct?
` A. Yes.
` Q. And a parity-check matrix doesn't
`necessarily determine whether a code will be
`systematic or not systematic, correct?
` MR. ROSATO: Objection, form,
` foundation.
` THE WITNESS: Correct.
`BY MR. GOLDENBERG:
` Q. But for any given LDPC parity-check
`matrix there exists a generator for that code
`that will produce a systematic code, correct?
` MR. ROSATO: Same objection.
` THE WITNESS: So if I understand
` that question, you're saying is there a way
` of setting up a generator that would make
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 29
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` a -- such a code systematic, right, that is
` potentially possible. Although, again,
` that doesn't mean that any specific low
` density parity check code or code
` implementation is itself systematic.
`BY MR. GOLDENBERG:
` Q. Well, what do you mean by
`implementation? The reason I ask is that part
`of the implementation for any parity-check
`matrix is picking a particular generator, so if
`that's what you mean by implementation -- is
`that what you meant by implementation?
` MR. ROSATO: Objection, form.
` THE WITNESS: I mean, I guess I take
` some issue with what you're saying, that an
` implementation requires picking a
` generator. You know, I think we've
` established that for low density
` parity-check matrix there exists
` generators, but I'm not clear that any
` given implementation has to utilize or have
` a generator matrix in order to construct or
` create code words.
`BY MR. GOLDENBERG:
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 30
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. I may have muddied the waters by
`guessing at what you meant by implementation.
` A. Okay.
` Q. And it seems like I guessed wrong.
`Let me just ask the question again.
` A. Okay.
` Q. For any given LDPC matrix there
`exists a generator matrix for the same code
`that is a systematic generator, correct?
` A. I believe that's the case for, you
`know, the general class of low density parity
`check codes that are -- that have valid sense
`of constraints, yeah.
` Q. Do you have the LUBY98 -- you don't
`have the prior references in your binder there,
`do you?
` A. No.
` MR. GOLDENBERG: Can I have your
` copy of the exhibit back. I just want to
` make a note on it.
` MR. ROSATO: Counsel, excuse me, I
` don't mean to interrupt.
` MR. GOLDENBERG: Let's go off the
` record.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 31
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` MR. ROSATO: Okay.
` (Discussion off the record.)
` (Exhibit Apple 1004 for
` IPR2017-00210, marked.)
`BY MR. GOLDENBERG:
` Q. Dr. Mitzenmacher, you've been handed
`an exhibit. It's marked Apple 1004, and it
`also says IPR2017-00210. Do you see that?
` A. Yes.
` Q. And do you recognize -- let me pause
`for a second.
` This exhibit that you have in front
`of you has been filed in multiple IPRs. It may
`have different exhibit numbers of the different
`six IPRs, but at least for IPR2017-00210, this
`document was Exhibit 1004. Do you understand?
` A. Okay.
` Q. And do you recognize the document
`you have in front of you?
` A. Yeah, it appears to be the --
`looking at the table of content for, "The
`Thirtieth Annual ACM Symposium on Theory of
`Computing," and then a paper on that conference
`which I am a coauthor.
`
`TSG Reporting - Worldwide 877-702-9580
`
`
`
`Page 32
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`
` Michael Mitzenmacher, PhD
` Q. Can we refer today to this exhibit
`that was Exhibit 1004 and IPR2017-00210, can we
`refer to that document as LUBY98?
` A. All right.
` Q. You'll know what I mean when I say
`LUBY98?
` A. I'll try to keep that in mind.
` Q. So Luby98 describes some LDPC codes,
`correct?
` A. Yes.
` Q. Every one of the codes described in
`LUBY98 has a corresponding systematic generator
`matrix, correct?
` A. We don't describe general matrices.
`There are various issues with your question.
`First, we tend to define families of code,
`although we did do experiments on some specific
`codes, so I think that would apply only to the
`extent that you're talking about the specific
`codes that we tested, but also we don't -- to
`the extent it applies, it would apply only to
`those and certainly not to the families that
`we're describing. A