UNITED STATES PATENT AND TRADEMARK OFFICE
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`In re: Inter Partes Review of:
`
`U.S. Pat. No. 7,116,710
`
`U.S. Pat. No. 7,421,032
`
`U.S. Pat. No. 7,421,781
`
`and U.S. Pat. No. 8,284,833
`
`Inventor: Hui Jin, et al
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`:
`
`:
`
`:
`
`:
`
`:
`
`:
`
`
`
`IPR No. Unassigned
`
`Assignee: California Institute of Technology
`
`
`Common Title: Serial Concatenation of Interleaved Convolutional Codes
`Forming Turbo-Like Codes
`
`
`
`
`
`Mail Stop PATENT BOARD
`Patent Trial and Appeal Board
`U.S. Patent and Trademark Office
`P.O. Box 1450
`Alexandria, Virginia 22313-1450
`
`
`Submitted Electronically via the Patent Review Processing System
`
`
`
`DECLARATION OF BRENDAN J. FREY
`
`1
`
`Hughes, Exh. 1059, p. 1
`
`

`

`Declaration of Brendan J. Frey
`
`
`I, Brendan J. Frey, declare as follows:
`
`1.
`
`I make this declaration based upon my own personal knowledge and,
`
`if called upon to testify, would testify competently to the matters contained herein.
`
`2.
`
`I have been informed that, for purposes of this declaration, the term
`
`“person of ordinary skill in the art” refers to person who, in May 18, 2001, had a
`
`formal educational level of at least a Ph.D. in a field related to coding or
`
`information theory. I have also been informed that the “person of ordinary skill in
`
`the art” would have been aware of all publications and other teachings related to
`
`the field of coding theory as of May 18, 2001.
`
`3.
`
`I am over the age of 18, have never been convicted of a felony or
`
`crime of moral turpitude and am legally competent to make this declaration.
`
`
`I.
`
`Background and Qualifications
`4. My qualifications are stated more fully in my curriculum vitae.
`
`Attached herewith. Here I provide a brief summary of my qualifications:
`
`5.
`
`I am currently a Professor at the University of Toronto in the
`
`Department of Electrical and Computer Engineering, and the Banting and Best
`
`Department of Medical Research, and the Department of Computer Science.
`
`6.
`
`I hold a Bachelor of Science Degree in Electrical Engineering from
`
`Hughes, Exh. 1059, p. 2
`
`

`

`the University of Calgary, a Masters of Science Degree in Electrical and Computer
`
`Engineering from the University of Manitoba, and Ph.D. in Electrical and
`
`Computer Engineering from the University of Toronto.
`
`7.
`
`I have authored a book entitled “Graphical Models for Machine
`
`Learning and Digital Communication.” that is relevant to the subject matter of this
`
`case. In addition, I have authored or co-authored nearly 181 articles in peer-
`
`reviewed journals, conference proceedings, texts, industry trade publications, and
`
`monographs.
`
`8.
`
`I have been asked to provide a statement of certain facts related to the
`
`inter partes review of the following U.S. Patents: i) U.S. Patent No. 7,116,710
`
`("the ’710 Patent"); ii) U.S. Patent No. 7,421,032 ("the ’032 Patent"); iii) U.S.
`
`Patent No. 7,421,781 ("the ’781 Patent"); and iv) U.S. Patent No. 8,284,833 ("the
`
`’833 Patent").
`
`9.
`
`I was an active contributor and collaborator in the community that
`
`included some of the inventors of the ‘710; ‘032; ‘781; and ‘833 patents around the
`
`time of the time of the alleged invention. In particular, I attended talks given by
`
`Dr. Robert McEliece and Dr. McEliece attended talks that I presented around the
`
`time of the alleged invention. These talks included the 1998 and 1999 Allerton
`
`Conferences held by the University of Illinois Urbana-Champaign in Allerton,
`
`Illinois, as well as the 2000 Brest 2nd International Symposium on Turbocodes and
`
`Hughes, Exh. 1059, p. 3
`
`

`

`Related Topics and the 2000 Sorrento ISIT conferences. Dr. McEliece, Dr.
`
`MacKay, and I attended and made presentations at the 1999 Institutive for
`
`Mathematics and its Applications (IMA) 1999 Summer Program: Codes, Systems
`
`and Graphical Models, which was held at the University of Minnesota on August
`
`2-13, 1999.
`
`10.
`
`I have been retained by the law firm of Baker Botts L.L.P., counsel
`
`for the petitioner Hughes Networks Systems, LLC and Hughes Communications,
`
`Inc. to provide my opinions as described below. For my efforts in connection with
`
`the preparation of this declaration I have been compensated at my standard rate of
`
`$500 per hour for this type of consulting activity. My compensation is in no way
`
`contingent on the results of these or any other proceedings relating to the above-
`
`captioned patent or matter.
`
`B. Publications, Presentations, and Software
`(i) “Irregular Turbocodes” - September 22-24, 1999 Allerton Conference
`
`11. Beginning in 1998, I collaborated with Dr. David MacKay to show
`
`that turbocodes could benefit from being made irregular codes in a similar way that
`
`Michael Luby, Michael Mitzenmacher, M. Amin Shokrollahi, Daniel A. Spielman,
`
`and others had adapted LDPC codes “irregular.” I used my software to simulate
`
`and test various irregular turbocodes. Additionally, I developed additional
`
`software to facilitate the simulation and testing of irregular turbocodes.
`
`Hughes, Exh. 1059, p. 4
`
`

`

`12. The result of that collaboration was a presentation at the 1999
`
`Allerton Conference on Communications, Control and Computing, and Computing
`
`in Allerton, Illinois (“1999 Allerton Conference”) in September 1999. The 1999
`
`Allerton Conference was held September 22-24, 1999. The 1999 Allerton
`
`Conference was open to the public for attendance. Any person who wanted to
`
`attend and was able to pay the attendance fee could attend the 1999 Allerton
`
`Conference. The 1999 Allerton Conference was considered one of two primary
`
`conferences on the topic of iterative decoding during the time. The 1999 Allerton
`
`conference was publicized and those who were interested in topics of iterative
`
`decoding were aware of the conference.
`
`13.
`
`I presented a series of slides on my work to show how a turbocodes
`
`could be made irregular and the benefits of making the turbocodes irregular. This
`
`presentation was entitled “Irregular Turbocodes” and I believe I presented on the
`
`afternoon of the first day of the conference on September 22, 1999. A copy of this
`
`slide presentation is shown at Exhibit 1034.
`
`14. After the 1999 Allerton conference, Dr. MacKay and I submitted a
`
`paper for publication in the Proceeding of the Thirty-Seventh Annual Allerton
`
`Conference on Communications, Control and Computing. Exhibit 1012 is a true
`
`and correct copy of the paper we submitted for publication. Dr. MacKay and I
`
`submitted the paper for publication on, or no later than October 1999. I believe the
`
`Hughes, Exh. 1059, p. 5
`
`

`

`“Irregular Turbocodes” paper as shown in Exhibit 1012 was on Dr. MacKay’s
`
`websites by October 1999--the month following my presentation. As I recall, Dr.
`
`MacKay was extremely prompt in posting all papers on his website immediately
`
`upon completion or submission for publication and that was the case with the
`
`“Irregular Turbocodes” paper. Like the talk, the paper was also entitled “Irregular
`
`Turbocodes” and contained the same figures presented at the 1999 Allerton
`
`Conference. The conference proceeding of the 1999 Allerton Conference,
`
`including my and Dr. MacKay’s paper entitled “Irregular Turbocodes” were later
`
`published by May 2000.
`
`(ii) “Irregular Turbocodes” - 2000 ISIT Conference
`
`15. Later, I attended the 2000 IEEE International Symposium on
`
`Information Theory (ISIT) conference in Sorrento, Italy. The 2000 ISIT
`
`conference was open to the public. Anyone who was interested and was willing to
`
`pay the entrance fee could attend the 2000 ISIT conference. The 2000 ISIT
`
`conference was considered one of two conferences on the topic of iterative
`
`decoding during the time. The ISIT conference covered a broader set of topics
`
`than the Allerton conference and drew a wider audience. The 2000 ISIT
`
`conference was publicized and those who were interested in topics of iterative
`
`decoding were aware of the conference.
`
`16. At the 2000 ISIT conference, on June 25, 2000, I presented a series of
`
`Hughes, Exh. 1059, p. 6
`
`

`

`slides on my work to show how a turbocodes could be made irregular and the
`
`benefits of making the turbocodes irregular. This presentation was entitled
`
`“Irregular Turbocodes.” A copy of this slide presentation is attached to this
`
`declaration as Exhibit 1035. The slides that I presented at the 2000 ISIT
`
`conference included the slides that I presented at the 1999 Allerton Conference
`
`with the addition of a single additional slide.
`
`17. Attendees
`
`to
`
`the 2000 ISIT conference received proceedings
`
`publication that contained a one-page paper for each of the talks presented at the
`
`2000 ISIT conference, including my and Dr. Mackay’s presentation on Irregular
`
`Turbocodes. The one-page publication that appeared in the Proceedings of the
`
`2000 ISIT conference was entitled “Irregular Turbocodes.” The full Proceedings
`
`of the 2000 ISIT conference were provided to attendees on arrival to the
`
`conference, which began on June 24, 2000. A true and correct copy of the 2000
`
`ISIT “Irregular Turbocodes” papers is shown at Exhibit 1048.
`
`(iii)
`“Irregular Turbo-Like Codes” - Second International
`Symposium on Turbocodes and Related Topics
`
`18.
`
`I also made a presentation on the topic of irregular turbocodes at the
`
`Second International Symposium on Turbocodes and Related Topics in Brest,
`
`France in September 2000 (“Brest 2000 Conference”). This presentation was
`
`entitled “Irregular Turbo-Like Codes.” A copy of this slide presentation is shown
`
`at Exhibit 1036. The slides that I presented at the Second International
`
`Hughes, Exh. 1059, p. 7
`
`

`

`Symposium on Turbocodes in Brest, France in September 2000 conference include
`
`the slides that I previously presented at the 2000 ISIT conference with the addition
`
`of five additional slides.
`
`19. Prior to the Brest 2000 Conference, Dr. Mackay and I submitted a
`
`paper for publication on the topic of my talk at the Brest 2000 Conference. The
`
`proceedings of the Brest conference, including my and Dr. MacKay’s paper
`
`entitled “Irregular Turbo-Like Codes” was distributed to the attendees at the
`
`conference.
`
`(iv)
`
`Contents of “Irregular Turbocodes”
`
`20. Below is a figure from slide 5 of the “Irregular Turbocodes”
`
`presentation to the 1999 Allerton Conference, the ISIT Sorrento conference and the
`
`Brest Turbocodes Conference:
`
`21. A person of ordinary skill in the art, upon seeing this diagram would
`
`
`
`Hughes, Exh. 1059, p. 8
`
`

`

`recognize that that f1 bits are repeated once, f2 bits are repeated twice (by “Rep 2”
`
`boxes), f3 bits are repeated three times (by “Rep 3” boxes), and fD bits are
`
`repeated D times (by “Rep D” boxes).
`
`22. With respect to the “permuter” element, it was my and Dr. MacKay’s
`
`understanding at the time of the “Irregular Turbocodes” presentations and papers
`
`that the term was interchangeable with the term “interleaver.” A person of
`
`ordinary skill in the art in May 18, 2001 would view the terms as interchangeable.
`
`23.
`
`In the structure labeled “Trellis representing constituent convolutional
`
`codes, average rate R,’” is a convolutional coder. Trellis constrained codes were
`
`explored in an earlier work by Dr. MacKay and me, entitled “Trellis Constrained
`
`Codes,” which was presented at the Thirty-Fifth Allerton Conference in 1997. A
`
`paper on the entitled “Trellis-Constrained Codes” was published in Proceeding of
`
`the Thirty-Fifth Allerton Conference. Ex. 1055. I would have expected those in
`
`attendance at my presentation on irregular turbocodes to at least have been aware
`
`of the earlier work on trellis-constrained codes. The version of this figure that
`
`appeared in the Proceedings of the 1999 Allerton Conference labeled this element
`
`as a “Convolutional code,” as shown below. This is consistent with the
`
`understanding of a person of ordinary skill in the art for the “Trellis representing
`
`constituent convolutional codes, average rate R,’” as shown in the slides from the
`
`presentation.
`
`Hughes, Exh. 1059, p. 9
`
`

`

`
`
`24.
`
`It was known at the time that a Repeat Accumulate (RA) code, as
`
`described by McEliece and others in their work “Coding Theorems for ‘Turbo-
`
`Like’ Codes” could be substituted for the ““Trellis representing constituent
`
`convolutional codes, average rate R’” or the “convolutional code” of the “Irregular
`
`Turbocodes” presentations and papers. A person of ordinary skill in the art would
`
`recognize that an accumulator of an RA code is a simple type of convolutional
`
`code. The RA code, as described by McEliece and others in their work “Coding
`
`Theorems for ‘Turbo-Like’ Codes” was therefore a simplified example of the
`
`““Trellis representing constituent convolutional codes, average rate R’” or the
`
`“convolutional code” in the “Irregular Turbocodes” presentations and papers.
`
`25. The equations that include the code rate in the “Irregular Turbocodes”
`
`presentations and papers apply to codes of different rates, including those with a
`
`rate that is close to one. A person of ordinary skill in the art would recognize that
`
`Hughes, Exh. 1059, p. 10
`
`

`

`the rate of a code, including the rate of the codes described in “Irregular
`
`Turbocodes” can be adjusted, based on the needs and limitations of the system in
`
`which the codes are used.
`
`26. A person of ordinary skill in the art would recognize that RA codes
`
`are in the same class of codes as the turbocodes of “Irregular Turbocodes.” In
`
`particular, the RA is a simple type of the convolutional code of “Irregular
`
`Turbocodes.” Because a Repeat Accumulate (RA) code is a type of convolutional
`
`code, the substitution of RA for the convolutional coder in “Irregular Turbocodes”
`
`would be straight forward to a person of ordinary skill in the art in May 18, 2000.
`
`Dr. MacKay and my work on “Irregular Turbocodes” demonstrated that making
`
`the turbocode irregular improved the performance of the coder. The performance
`
`of codes is typically measured as a function of Bit Error Rate or Word Error Rate
`
`versus the energy per bit to noise power spectral density ratio (Eb/No) (dB). As
`
`shown, for example, in slides 11 and 12 of the “Irregular Turbocodes” presentation
`
`from the 1999 Allerton Conference, the presentation demonstrated that making the
`
`turbocode irregular improved the turbocode’s performance. Because an RA code
`
`is a simple kind of turbo code, a person of ordinary skill in the art would expect to
`
`achieve a similar performance improvement by making an RA code irregular.
`
`27. A person of ordinary skill in the art, reading “Irregular Turbocodes”
`
`would understand that the block labeled “convolutional coder” would be
`
`Hughes, Exh. 1059, p. 11
`
`

`

`implemented using a mod-2 or exclusive-OR sum, because convolutional
`
`operations on bits are all performed using mod-2 or exclusive-OR sums.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Hughes, Exh. 1059, p. 12
`
`

`

`I hereby declare that all statements made herein of my own knowledge are
`
`true and that all statements made on information or belief are believed to be true;
`
`and fiirther that these statements were made with the knowledge that willful false
`
`statements and the like so made are punishable by fine or imprisonment, or both,
`
`under Section 1001 of Title 18 of the United States Code.
`
`Dated: 0619 /‘/ 201%
`______fiL___
`
`Brendan J. Frey
`
`Hughes,Exh.1059,p.13
`
`Hughes, Exh. 1059, p. 13
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae – Summary
`
`October 2, 2014
`
`Born August 29, 1968, Calgary, Alberta, Canada. Canadian citizen.
`Tel: +1 416 978 7001; Fax: +1 416 978 4425. E-mail: frey@psi.toronto.edu; Web: www.psi.toronto.edu.
`Mail: University of Toronto, 10 King’s College Road, Toronto, Ontario, Canada, M5S 3G4.
`Higher education
`Ph.D. 1997, Univ of Toronto, Elec and Comp Eng. Supervisor: G.E. Hinton (co-supervisor G. Gulak)
`M.Sc. 1993, Univ of Manitoba, Elec and Comp Eng. Supervisor: H.C. Card.
`B.Sc. (Honours) 1990, Univ of Calgary, Elec Eng.
`Current appointments
`Professor, University of Toronto, Canada, Electrical and Computer Engineering
`Cross-appointed Professor, Banting and Best Department of Medical Research, and Computer Science
`Fellow, Canadian Institute for Advanced Research
`Senior Fellow of Massey College
`Distinctions
`John C. Polanyi Award, NSERC (2012)
`EWR Steacie Fellow, NSERC (2009)
`Fellow of the AAAS (2009)
`Fellow of the IEEE (2008)
`Canada’s Top 40 Under 40 Award, Caldwell Partners (2007)
`Canada Research Chair, Information Processing and Machine Learning (2007-2012)
`Conference Paper Awards, UAI (2008, 2011), CVPR (2005)
`Fellow of the Canadian Institute for Advanced Research (2006-2016)
`Premier’s Research Excellence Award, Province of Ontario (2000-2005)
`Undergraduate Teaching Awards, University of Toronto (2006, 2007)
`Beckman Fellow, University of Illinois at Urbana-Champaign (1997-1999)
`Supervision
`Current: 3 PDF, 6 PhD, 2 MSc, 3 RA. Graduated: 7 PDF, 18 PhD, 9 MSc, 1 RA. Graduate thesis committees:
`30 (excluding my own students).
`Selected professional activities
`Technical Advisory Board, Microsoft Research India, 2010-present. Tutorial Chair, ISIT 2008. Co-Chair,
`AISTATS 2003. Assoc Ed, IEEE Trans PAMI 2001-2005. Guest Co-Ed-in-Chief, IEEE Trans IT 2002. Co-
`Chair, Canadian Work Info Theory 2003. Grant Review Panels: Canada Research Chairs; NSERC; NSF;
`ISF; SFI; Sloan Foundation. Peer reviews: BMC Bioinf; Cell; CVPR; ICCV; ICML; IEEE Trans IP, IT, NN,
`PAMI, SAP; IEEE CL, JSAC; ISMB; JACM; JASA; JMLR; Nature; Nat Gen; Nat Meth; Neur Comp; NIPS;
`PNAS; RECOMB; Science; UAI.
`Research grants and unrestricted industrial gifts in past five years
`NSERC JC Polanyi Award
`PI
`2012-2014
`CIHR Research Grant
`PI
`2010-2014
`NSERC EWR Steacie Grant
`PI
`2009-2011
`Canadian Institute for Advanced Research PI
`2006-2010
`NSERC Discovery Grants
`PI
`1999-2017
`Microsoft Research Excellence Award
`PI
`2001-2008
`CIHR Research Grant
`PI
`2006-2009
`Genome Canada Sub-Award
`PI
`2006-2011
`
`$250,000
`$520,000
`$364,000
`$120,000
`$936,000
`$390,000
`$258,880
`$645,000
`
`1
`
`Hughes, Exh. 1059, p. 14
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`1. Professional Appointments
`
`Jul. 2007 – present:
`
`Full Professor (tenured)
`University of Toronto, Canada
`Department of Electrical and Computer Engineering
`
`Jan. 2010 – present:
`
`Technical Advisory Board, Microsoft Research India
`
`Jul. 2011 – present:
`
`Senior Fellow, Massey College, University of Toronto, Canada
`
`Aug. 2008 – Jul. 2009:
`
`Visiting Researcher
`Microsoft Research, Cambridge, United Kingdom
`Machine Learning and Perception Group
`
`Sep. 2008 – Aug. 2009: Visiting Professor
`University of Cambridge, United Kingdom
`Cavendish Laboratories and Darwin College
`
`Mar. 2005 – present:
`
`Cross-Appointed Professor
`University of Toronto, Canada
`Banting and Best Department of Medical Research
`
`Mar. 2005 – present:
`
`Cross-Appointed Professor
`University of Toronto, Canada, Department of Computer Science
`
`Feb. 2000 – present:
`
`Research Management Consultant, Microsoft Research, Redmond, USA
`
`Jul. 2004 – Jun. 2007:
`
`Associate Professor (tenured)
`University of Toronto, Canada, Electrical and Computer Engineering
`
`Jul. 1998 – Dec. 2006:
`
`Adjunct Professor
`University of Illinois at Urbana-Champaign, USA
`Electrical and Computer Engineering
`
`Jul. 2001 – Jun. 2004:
`
`Assistant Professor
`University of Toronto, Canada, Electrical and Computer Engineering
`
`Dec. 1998 – Jun. 2001:
`
`Assistant Professor
`University of Waterloo, Computer Science
`
`Jul. 1997 – Jan. 1999:
`
`Beckman Fellow
`University of Illinois at Urbana-Champaign, USA
`Beckman Institute for Advanced Science and Technology
`
`Nov. 1997 – Dec. 1997:
`
`Invited Researcher
`Cambridge University, England
`Isaac Newton Institute for the Mathematical Sciences
`
`Sep. 1990 – Sep. 1991:
`
`Junior Research Scientist
`Bell-Northern Research, Advanced Development Group
`
`2
`
`Hughes, Exh. 1059, p. 15
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`2. Most significant research contributions
`
`2.1 DECIPHERING THE SPLICING CODE
`Y Barash, J Calarco, W Gao, Q Pan, X Wang, O Shai, BJ Blencowe, BJ Frey1. Nature 456, 53-59, May 2010.
`
`Alternative splicing has a crucial role in the generation of biological complexity, and its misregulation is
`often involved in human disease. We described the assembly of a splicing code, which uses combinations
`of hundreds of RNA features to predict tissue-dependent changes in alternative splicing for thousands
`of exons. The code determines new classes of splicing patterns, identifies distinct regulatory programs
`in different tissues, and identifies mutation-verified regulatory sequences. Widespread regulatory strate-
`gies are revealed, including the use of unexpectedly large combinations of features, the establishment
`of low exon inclusion levels that are overcome by features in specific tissues, the appearance of features
`deeper into introns than previously appreciated, and the modulation of splice variant levels by transcript
`structure characteristics. The code detected a class of exons whose inclusion silences expression in adult
`tissues by activating nonsense-mediated messenger RNA decay, but whose exclusion promotes expres-
`sion during embryogenesis. The code facilitates the discovery and detailed characterization of regulated
`alternative splicing events on a genome-wide scale. The online web tool http://genes.toronto.edu/wasp
`has been accessed over 50,000 times.
`
`Citations (past 22 months): 65 Thomson ISI, 117 GoogleScholar.
`
`2.2 THE AFFINITY PROPAGATION ALGORITHM
`
`BJ Frey and D Dueck. Science 315, 972-976, February 2007.
`
`Summarizing data by identifying a subset of representative examples is important for scientific data anal-
`ysis and in engineered systems. Such ‘exemplars’ can be found by randomly choosing an initial subset
`of data points and then iteratively refining it, but this only works well if that initial choice is close to a
`good solution. We described a new method called ‘affinity propagation’, which takes as input measures
`of similarity between pairs of data points. Real-valued messages are exchanged between data points until
`a high-quality set of exemplars and corresponding clusters gradually emerges. Because of its simplic-
`ity, general applicability, and performance, affinity propagation could be of broad value in science and
`engineering. A Google search for “affinity propagation” returns over 70,000 hits, and in the past year,
`the affinity propagation web tool http://www.psi.toronto.edu/affinitypropagation was accessed over
`500, 000 times.
`
`Citations (past 5 years): 410 Thomson ISI, 1011 GoogleScholar.
`
`2.3 PROFILING ALTERNATIVE SPLICING IN MAMMALIAN TISSUES
`
`Q Pan, O Shai, LJ Lee, BJ Frey, BJ Blencowe, Nature Genetics 40, 1413-1415, 2008; Q Pan, O Shai, C Mis-
`quitta, W Zhang, N Mohammed, T Babak, H Siu, TR Hughes, Q Morris, BJ Frey, BJ Blencowe, Molecular
`Cell 16, 929-941, 2004.
`
`In this work, we developed computational techniques to detect biologically-significant patterns of alter-
`native splicing, using DNA microarray experiments and second generation RNA sequencing (RNA-Seq).
`Microarrays and RNA-Seq are used to detect levels of expressed transcripts in diverse cell types or tissue
`samples. In contrast to low-throughput laboratory techniques, these methods can measure the levels of
`thousands to millions of transcripts at once. The computational techniques developed in my group were
`used to analyze microarray data and more recently RNA-Seq data collected by my collaborators, Benjamin
`
`1Bold indicates BJ Frey is a lead/corresponding author. Trainees of BJ Frey are underlined. Note that in scientific publications,
`the names of principal investigators often appear last in author lists. Trainees of BJ Frey are underlined
`
`3
`
`Hughes, Exh. 1059, p. 16
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`J. Blencowe and Timothy R. Hughes. Our techniques explain the measured transcript levels using hidden
`variables that have biological significance. Using these techniques, we were able to: reveal genome-wide
`trends in tissue-dependent alternative splicing of mammalian gene variants, show that transcription and
`alternative splicing act independently to determine gene function, and identify some of the genetic deter-
`minants of alternative splicing. These data were used in subsequent research to infer the ‘splicing code’
`(see above).
`
`Citations (past 6 years): 559 Thomson ISI, 742 GoogleScholar.
`
`2.4 FACTOR GRAPHS AND THE SUM-PRODUCT ALGORITHM
`
`BJ Frey, Graphical Models for Machine Learning and Digital Communication, MIT Press, 1998. FR Kschis-
`chang, BJ Frey, H-A Loeliger, IEEE Trans Info Theory 47:2, 498-519, 2001.
`
`Graph-based models and their associated distributed algorithms for inference and learning are playing
`an increasingly important role in science and engineering. In this work, we describe fundamental con-
`nections between algorithms introduced in the information theory and artificial intelligence communities
`and introduce the “factor graph”, which is a network description of a system of random variables and
`which offers advantages over Bayesian networks and Markov random fields. Factor graphs (now in-
`cluded in at least 10 university textbooks, including CM Bishop 2006, R Blahut 2003, MI Jordan 2006)
`have become popular for describing the sum-product (aka “loopy belief propagation”) generalization of
`dynamic programming, and have been used to develop methods for visual data analysis; to solve random
`satisfiability problems that are orders of magnitude larger than previously solvable problems (Mezard et
`al, Science 2002); and to infer bio-molecule interaction networks (Yeang et al, JCB 2004).
`
`Citations (past 12 years): 1562 Thomson ISI, 3377 GoogleScholar.
`
`2.5 THE WAKE-SLEEP ALGORITHM FOR UNSUPERVISED NEURAL NETWORKS
`
`GE Hinton, P Dayan, BJ Frey, RM Neal. Science 268, 1158-1161, 1995.
`
`Most engineering systems and models of perception are feed-foward – they try to directly predict class
`membership or the target value using sensory data as input. In work that was reported in The Daily Tele-
`graph in May 1995 and broadcast on The Discovery Channel in June 1995, we introduced a machine learning
`algorithm that can jointly train a feed-back (generative) network to model an unknown stochastic process
`that can synthesize perceptual data and train a feed-forward (recognition) network to invert the feed-back
`network, i.e., predict class membership from sensory input. This technique enables the learning of feed-
`forward systems that are much more complex than linear systems or low-order systems, and has been
`used as a model of visual perception and to design engineering systems for pattern classification.
`
`Citations (past 17 years): 190 ISI, 462 GoogleScholar.
`
`4
`
`Hughes, Exh. 1059, p. 17
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`3. Distinctions
`• John C. Polanyi Award
`NSERC, Canada, 2012
`• Senior Fellow of Massey College
`University of Toronto, Canada, 2011
`• EWR Steacie Fellow
`NSERC, Canada, 2009
`• Technical Advisory Board
`Microsoft Research India, 2010-present
`• Fellow of the IEEE
`IEEE, USA, December 2008
`• Fellow of the AAAS
`AAAS, USA, January 2009
`• Invited Speaker, ideaCity
`Toronto, Canada, June 2008
`• Canada’s Top 40 Under 40
`Caldwell Partners, Canada, May 2007
`• Fellow of the Canadian Institute for Advanced Research
`Program on Genetic Networks, 2006-2011
`• Canada Research Chair
`Information Processing and Machine Learning, 2007-2012
`• Outstanding Undergraduate Teaching Award
`Department of Electrical and Computer Engineering, Faculty of Applied Sciences and Engineering
`University of Toronto, 2007
`• Plenary Lecture, International Symposium on Information Theory
`Seattle, WA, July 2006
`• Fellow of the Canadian Institute for Advanced Research
`Program on Neural Computation, 2006-2013
`• Plenary Lecture, Machine Learning Workshop
`Snowbird, UT, April 2006
`• Outstanding Undergraduate Teaching Award
`Department of Electrical and Computer Engineering, Faculty of Applied Sciences and Engineering
`University of Toronto, 2006
`• Runner-Up Best Paper Award
`IEEE Conference on Computer Vision and Pattern Recognition, 2005
`• Associate of the Canadian Institute for Advanced Research
`Program on Neural Computation, 2004-2006
`• Premier’s Research Excellence Award
`Province of Ontario, 2000-2005
`
`5
`
`Hughes, Exh. 1059, p. 18
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`• Fellow of the Beckman Institute for Advanced Science and Technology
`Beckman Foundation, 1997-1999
`• Invited Researcher of the Isaac Newton Institute for Mathematical Sciences
`Cambridge, England, 1997
`• NSERC 1967 Science and Engineering Award, 1993-1997
`
`4. Reported in the News
`• BBC Radio The Naked Scientist, Spring 2010
`• CBC Radio Here and Now, May 7, 2010
`• Toronto Star, May 6, 2010, main headline of front page, “U of T cracks the code”, Joseph Hall
`• Globe and Mail, May 7, 2010, “With fewer genes than a block of wood, humans rely on a hidden
`code”, Paul Taylor
`• Nature Magazine News and Views, May 6, 2010, “The code within the code”
`• Nature Magazine Editorial, May 6, 2010
`• Vancouver Sun, May 6, 2010, “Researchers crack hidden ‘splicing code’ in genes”
`• Stanford Biomedical Computation Review, Summer 2007, “Clustering without limits”, Kathy Miller
`• Canadian Technology News, June 8, 2007
`• Globe and Mail, May 8, 2007, “Forty talents in bloom”, Peter Evans
`• Science Magazine, February 16, 2007, “Perspective: Where Are the Exemplars?”, Marc Mezard
`• Backbone Magazine, May 1, 2007, “Computing on a new level”, Ian Harvey
`• ACM TechNews, March 3, 2007, “Canadian researcher proposes algorithm to match data”, Shane
`Schick
`• Computing Canada, March 2, 2007, “Machine learning algorithm finds affinity in data”
`• Technology Trends, March 5, 2007, “Affinity propagation for faster computing”, Roland Piquepaille
`• ACM TechNews, February 28, 2007, “Canadian researcher proposes algorithm to match data”,
`Shane Schick
`• Medical News Today, September 4, 2005, “U of T researcher uncovers genetic instructions to build
`life”
`• Canadian Institutes of Health Research News, Dec 19, 2005, “Artificial intelligence program unlocks
`gene secrets”
`• ACM TechNews, August 31, 2005, “Timely topics for IT professionals”
`• Science and Theology News, September 6, 2005, “Life or something like it”
`• Medical News Today, December 22, 2004, “New technique provides insights into gene regulation”
`
`6
`
`Hughes, Exh. 1059, p. 19
`
`

`

`Brendan J. Frey
`
`Curriculum Vitae
`
`October 2, 2014
`
`• BioMed Central’s Most-Viewed Article Over the Last 30 Days, January 2005 (ranked as the #1 article
`by the Faculty of 1000)
`• Beckman Institute Annual Newsletter, January 15, 2001, “Like Magic: Beckman Institute Fellow
`Brendan Frey”
`• Toronto Star, January 13, 1998, “Why dreams help us to make sense of the world”
`
`5. Research Grants
`
`PI indicates Principal Investigator; co-PI indicates somebody else is the PI, but I wrote a portion of the
`grant application and part of the budget is under my control.
`• NSERC Discovery Grant and Accelerator, Apr. 2013 - Mar. 2017, $368,000 ($92,000/yr)
`A Unified Model of Gene Regulation
`B. J. Frey (PI)
`• OGI Spark Grant, Sep. 2011 - Aug. 2012, $50,000 ($50,000/yr)
`First-of-a-Kind Web Tool for Exploring Splicing Misregulation in Human Disease
`B. J. Frey (PI) and B. J. Blencowe
`• NSERC, Sep. 2012 - Aug. 2014, $250,000 ($125,000/yr)
`John C. Polanyi Award
`B. J. Frey (PI) and B. J. Blencowe
`• NSERC, Sep. 2009 - Aug. 2011, $375,000 ($187,500/yr)
`EWR Steacie Fellow Grant
`B. J. Frey (PI)
`• CIHR Operating Grant, Sep. 2010 - Aug. 2014, $520,000 ($130,000/yr)
`Genome-wide analysis of the regulation of alternative splicing
`B. J. Frey (PI)
`• Canadian Institute for Advanced Research, Jul. 2006 - Jul. 2009, $75,000 ($25,000/yr)
`Programs on Neural Computation and Adaptive Perception and Genetic Networks
`• NSERC Discovery Grant, Apr. 2008 - Mar. 2013, $260,000 ($52,000/yr)
`Affinity Propagation: Theory and Practice
`B. J. Frey (PI)
`• CIHR Operating Grant, Apr. 2006 - Apr. 2009, $258,000 ($86,000/yr)
`Large-Scale Discovery of Mammalian Gene Variants and Their Regulatory Motifs
`B. J. Frey (PI)
`• Genome Canada Grant, Dec. 2006 - Nov. 2010, $645,000 (my sub-award) ($161,250/yr)
`Structural and Functional Annotation of the Human Genome For Disease Study
`B. J. Blencowe (co-PI), B. J. Frey (co-PI), R. Hegele (PI), T. R. Hughes (co-PI), S. Scherer (co-PI)
`• NSERC Discovery Grant, Apr. 2003 - Mar. 2008, $230,000 ($46,000/yr)
`Learning Large-Scale Generative Models for Video Analysis
`B. J. Frey (PI)
`• Microsoft Research Gift, Sep. 2001 - Dec. 2007, $390,000 ($65,000/yr)
`In Recognition of Research Excellence in Vision and Computational Biology Research
`B. J. Frey (PI)
`
`7
`
`Hughes, Exh. 1059, p. 20
`
`

`

`Brendan J.