`
`______________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`TOYOTA MOTOR CORPORATION
`
`Petitioner
`
`v.
`
`AMERICAN VEHICULAR SCIENCES
`
`Patent Owner
`
`Patent No. 5,845,000
`
`Issue Date: December 1, 1998
`
`Title: OPTICAL IDENTIFICATION AND MONITORING SYSTEM USING
`PATTERN RECOGNITION FOR USE WITH VEHICLES
`
`DECLARATION OF CRIS KOUTSOUGERAS, PH.D. IN SUPPORT OF
`AVS’S RESPONSE UNDER 37 CFR § 42.120
`
`Case No. IPR2013-00424
`
`AVS EXHIBIT 2002
`Toyota Inc. v. American Vehicular Sciences LLC
`IPR2013-00424
`
`
`
`I.
`
`INTRODUCTION AND SUMMARY OF OPINIONS
`
`1.
`
`My name is Cris Koutsougeras. I am a professor at the Department of
`
`Computer Science and Industrial Technology at Southeastern Louisiana University
`
`in Hammond, Louisiana, where I teach courses in Computer Science and in
`
`Engineering Technology. My background consists of degrees in Electrical
`
`Engineering, Computer Engineering, and Computer Science. Of my past work,
`
`pertinent to the present review is my research on neural networks, robotics control
`
`and intelligence, and sensors and their interfacing.
`
`2.
`
`I have been hired by American Vehicular Sciences (“AVS”) in
`
`connection with the above-captioned Inter Partes Reexamination Proceeding
`
`(“IPR”) before the United States Patent and Trademark Office. In the below
`
`paragraphs, I provide my opinion that at least claims 10, 11, 16, 17, 19, 20, and 23
`
`of U.S. Patent No. 5,845,000 (“the ‘000 patent”) at issue in the IPR are not
`
`anticipated or obvious in view of the grounds for review.
`
`II.
`
`PROFESSIONAL BACKGROUND AND QUALIFICATIONS
`
`3.
`
`My background consists of a B.S. degree in Electrical Engineering, an
`
`M.S. degree in Computer Engineering, and a Ph.D. degree in Computer Science.
`
`4.
`
`I received my B.S. degree in 1983 from the National Technical
`
`University of Athens, my M.S. degree in 1984 from the University of Cincinnati,
`
`and my Ph.D. in 1988 from Case Western Reserve University. My Ph.D. research
`
`1
`
`
`
`and dissertation was on the topic of neural networks and, more specifically, on
`
`algorithms for training feed-forward types of neural networks.
`
`5.
`
`I also have experience in automotive technology involving external
`
`object detection systems, collision warning systems, and the use of pattern
`
`recognition technology in such systems, as I have participated in the DARPA 2005
`
`Grand Challenge competition with a team that built an autonomous vehicle
`
`designed to drive completely unassisted in unknown and unrehearsed cross-country
`
`environments. The vehicle was a regular production SUV that was modified to be
`
`controlled by computers aided by sensors including Ladars and GPS.
`
`6.
`
`I am a professor at the Department of Computer Science and
`
`Industrial Technology at Southeastern Louisiana University in Hammond,
`
`Louisiana, and teach courses in Computer Science and in Engineering Technology.
`
`I served as department head of that department from 2006 to 2011.
`
`7.
`
`Prior to joining Southeastern Louisiana University, I was a faculty
`
`member of the Department of Electrical Engineering and Computer Science at
`
`Tulane University in New Orleans, Louisiana, from 1988 to 2006.
`
`8.
`
`A more detailed account of my work experience, qualifications, and
`
`list of publications is included in my Curriculum Vitae, which is attached to this
`
`Declaration.
`
`2
`
`
`
`III. COMPENSATION AND MATERIALS CONSIDERED
`
`9.
`
`I am being compensated for my time as an expert witness on this
`
`matter at $260 per hour. My compensation, however, does not depend in any way
`
`on my opinions or conclusions, nor on the result of this proceeding.
`
`10.
`
`11.
`
`12.
`
`I am not an employee of AVS or any affiliate, parent, or subsidiary.
`
`I have not served as an expert in the last 10 years.
`
`In arriving at my opinions, I considered the following documents:
`
` U.S. Pat. No. 5,845,000;
`
` Prosecution History of U.S. Pat. No. 5,845,000;
`
` The Patent Trial and Appeals Board’s Decision to Institute Inter
`
`Partes Review;
`
` Toyota’s Petition for Inter Partes Review;
`
` The Declaration of Dr. Nikolaos Papanikolopoulos;
`
` The Transcript of the Deposition of Dr. Papanikolopoulos
`
` U.S. Pat. No. 6,553,130 to Lemelson;
`
` U.S. Pat. No. 5,214,408 to Asayama;
`
` Pomerleau, Dean, “Neural Networking Perception for Mobile
`
`Robot Guidance,” CMU-CS-92-115, AD-A249927, Feb. 16,
`
`1992;
`
`3
`
`
`
` Japanese Unexamined Patent Application Publication JP-H06-
`
`267303 to Mizukoshi;
`
` Japanese Unexamined Patent Application Publication JP-S62-
`
`131837 to Yanagawa; and
`
` The additional patents and references I cite in this declaration in
`
`support of my opinions.
`
`IV. OVERVIEW OF THE ‘000 PATENT
`
`A.
`
`13.
`
`Technical Overview of the ‘000 Patent
`
`The ‘000 patent relates, in relevant part, to a system for monitoring at
`
`least one object exterior to a vehicle and to a headlight dimming system. In
`
`particular, the ‘000 patent involves identifying objects or radiation sources outside
`
`the vehicle, and affecting other systems in the vehicle in response to the
`
`identification.
`
`14.
`
`Each of the claims at issue in the IPR requires at least one of the
`
`following specific features: claims 10, 11 and 19 require using a “trained pattern
`
`recognition means” that is “structured and arranged to apply a pattern recognition
`
`algorithm generated from data of possible exterior objects and patterns of received
`
`electromagnetic illumination from the possible exterior objects”; claims 16, 17,
`
`and 20 require “trained pattern recognition means” that is “structured and arranged
`
`to apply a pattern recognition algorithm generated from data of possible sources of
`
`4
`
`
`
`radiation including lights of vehicles and patterns of received radiation from the
`
`possible sources”; and claim 23 requires “generating a pattern recognition
`
`algorithm from data of possible exterior objects and patterns of received
`
`electromagnetic illumination from the possible exterior objects.”
`
`15. With respect to claims 10, 11, 16, 17, 19, 20, and 23, the ‘000
`
`patent’s pattern recognition algorithm is trained with (1) data of possible exterior
`
`objects (claims 10, 11, 19 and 23) or data of possible radiation sources (claims 16,
`
`17, and 20) and (2) patterns of received waves (e.g., patterns of received
`
`electromagnetic illumination from the possible exterior objects (claims 10, 11, 19,
`
`and 23) or patterns of received radiation from the possible sources (claims 16, 17,
`
`and 20)). (See, e.g., the ‘000 patent at 16:13-17:21 and 19:26-40.) In other words,
`
`the ‘000 patent’s pattern recognition sub-system is trained to recognize how waves
`
`behave when they are received from a given object or radiation source. A trained
`
`pattern recognition model, such as a neural network is very different from a
`
`traditional program-based model (otherwise known as a “symbolic” approach). A
`
`computer program can be used if there exists a well-understood process (recipe to
`
`make it plain) that is expressible in finite terms and which can be used to determine
`
`the output that corresponds to an input. In order for a programmer to produce a
`
`program, the programmer must know exactly this process and what part(s) or
`
`features of the input are relevant in this process. Sometimes, however, it can be
`
`5
`
`
`
`very difficult, if not impossible, to know this recognition or decision process, or it
`
`may not be expressible in finite terms, or it may not be known which input parts
`
`(variables) or combination of input variables are sufficient to uniquely and
`
`confidently determine the output. Then the alternative to traditional programming
`
`is to use trainable systems which essentially use statistical methods to interpolate
`
`from input-output example instances.
`
`16. What the ‘000 patent discloses is that waves received from objects
`
`(e.g., patterns of received electromagnetic illumination from the possible exterior
`
`objects (claims 10, 11, 19, and 23)) and waves received from radiation sources
`
`(e.g., patterns of received radiation from the possible sources (claims 16, 17, and
`
`20)) should carry enough information in their patterns to identify these objects and
`
`locations etc. But it is very difficult to isolate and extract this information by some
`
`stepwise process which will yield this identification in some finite steps. This is
`
`because we may not be able to express this process in finite terms, and/or because
`
`we do not know which parts of the waves (input) to use and how to combine them
`
`in order to determine the output. Therefore, the ‘000 patent discloses the use of
`
`neural networks, for example.
`
`17.
`
`In turn, neural networks require training to perform a task, using a
`
`certain set of inputs and the outputs that correspond to those inputs (these input-
`
`output pairs comprise the training set). The choice of training set is an important
`
`6
`
`
`
`key to the quality of the training of a neural net. With neural nets the developer
`
`does not need to encode the process which the system is expected to perform
`
`(because the system is expected to “learn” it). Instead, the developer “teaches” or
`
`“trains” the system what it is expected to perform by providing example input and
`
`output samples. But what the system will learn, or how well it will learn, depends
`
`on what was provided as the training set as well as what was chosen to provide as
`
`input. There is not a single, unique way or method known to choose the input
`
`features and training set that will guarantee the optimal training for any possible
`
`application. The inventor of the ‘000 patent considered that it was advantageous to
`
`use (1) data of possible exterior objects (claims 10, 11, 19 and 23) or data of
`
`possible radiation sources (claims 16, 17, and 20) and (2) patterns of received
`
`waves (e.g., patterns of received electromagnetic illumination from the possible
`
`exterior objects (claims 10, 11, 19, and 23) or patterns of received radiation from
`
`the possible sources (claims 16, 17, and 20)).
`
`18. Considering an object detection system, there might be an infinite
`
`number of possible input objects, when one factors in possible shapes, colors,
`
`angles, etc. In particular, in a vehicle-based system for detecting cars, it would be
`
`difficult to program such a system to compare a received image of a car to a
`
`database of images of all possible car models, in all possible colors, from all
`
`possible angles. Pattern recognition algorithms accommodate large variability in
`
`7
`
`
`
`possible targets. A pattern recognition system based on a neural network is
`
`different from a traditional computer program because it does not just compare a
`
`detected object to a database to find a match. For example, a pattern recognition
`
`algorithm calculates degrees of similarity between something it has been informed
`
`is a car, versus something it has been informed is not a car, and does so based on
`
`statistics extracted from the training set and reflected in its structure during
`
`training. The larger the training set, with more and balanced positive and negative
`
`examples that the system is given, the higher the degree of confidence that it will
`
`be properly trained to perform the intended function.
`
`19.
`
`The ‘000 patent discloses and claims a specific method for generating
`
`the “training set.” The ‘000 patent discloses training the algorithm with (1) data of
`
`possible exterior objects or data of possible radiation sources and (2) patterns of
`
`received waves (e.g., patterns of received electromagnetic illumination from the
`
`possible exterior objects or patterns of received radiation from the possible
`
`sources.) For example, if the vehicle uses a radar receiver, a neural network could
`
`be trained with examples of received radar waves from possible objects such as
`
`cars, plus labels to indicate the identification and possibly other information
`
`relating to the object. This “other information” might include, for example,
`
`location information. The statements “received waves from the possible exterior
`
`objects” and “received electromagnetic illumination from the possible exterior
`
`8
`
`
`
`objects” are understood to describe actual readings of real waves (e.g., real
`
`electromagnetic illumination) from actual possible exterior objects. The statement
`
`“received radiation from the possible sources” is understood to describe actual
`
`readings of real waves (e.g., real radiation) from actual possible radiation sources.
`
`20.
`
`In the case of a radar system, the examples of received radar waves
`
`from possible objects used to train the system are real radar waves or based on real
`
`radar waves, so that the system can recognize radar waves received from that same
`
`object or a similar one when the vehicle is later driving in actual conditions. This
`
`can be done, for example, by putting actual examples of a possible object in front
`
`of a vehicle radar system, subjecting the object with radar waves that are received
`
`back by the system, and then labeling the object for the system. This is different
`
`from other possible ways to train a pattern recognition system, such as through
`
`completely simulated data, which I discuss below.
`
`21.
`
`I also considered the prosecution history of the ‘000 patent. The
`
`prosecution history of the ‘000 patent did not involve any arguments or disclaimers
`
`that were relevant to my opinions in this declaration.
`
`B.
`
`22.
`
`Claim Construction
`
`I understand that the first step in any invalidity analysis is to construe
`
`the meaning of the claims. The Board has done so in its Decision Instituting Inter
`
`Partes Review. In particular, the Board made the following claim constructions:
`
`9
`
`
`
`“pattern recognition algorithm” The Board defined this as “an
`
`23.
`
`algorithm which processes a signal that is generated by an object, or is modified by
`
`interacting with an object, for determining to which one of a set of classes the
`
`object belongs.”
`
`24.
`
`“trained pattern recognition means” The Board defined this as “a
`
`neural computer or microprocessor trained for pattern recognition, and equivalents
`
`thereof.”
`
`25.
`
`“identify” and “identification”
`
` The Board defined
`
`this as
`
`“determining that the object belongs to a particular set or class.”
`
`26.
`
`“transmitter means
`
`for
`
`transmitting electromagnetic waves
`
`to
`
`illuminate the at least one exterior object” The Board defined this as a transmitter,
`
`which covers infrared, radar, and pulsed GaAs laser systems and transmitters
`
`which emit visible light.
`
`27.
`
`“reception means for receiving reflected electromagnetic illumination
`
`from the at least one exterior object” and “reception means for receiving
`
`electromagnetic radiation from the exterior of the vehicle” The Board defined this
`
`as a CCD array and CCD transducer.
`
`28.
`
`“processor means coupled to said reception means for processing
`
`said received illumination and creating an electronic signal characteristic of said
`
`exterior object based thereon” and “processor means coupled to said reception
`
`10
`
`
`
`means for processing the received radiation and creating an electronic signal
`
`characteristic of the received radiation” The Board defined this as a processor.
`
`29.
`
`“categorization means coupled
`
`to said processor means
`
`for
`
`categorizing said electronic signal
`
`to
`
`identify said exterior object, said
`
`categorization means comprising trained pattern recognition means” and
`
`“categorization means coupled to said processor means for categorizing said
`
`electronic signal to identify a source of the radiation, said categorization means
`
`comprising trained pattern recognition means” The Board defined this as a neural
`
`computer, a microprocessor, and their equivalents.
`
`30.
`
`“output means coupled to said categorization means for affecting
`
`another system in the vehicle in response to the identification of said exterior
`
`object” and “output means coupled to said categorization means for dimming the
`
`headlights in said vehicle in response to the identification of the source of the
`
`radiation” The Board defined this as an electronic circuit or circuits capable of
`
`outputting a signal to another vehicle system.
`
`31.
`
`“measurement means for measuring the distance from the at least one
`
`exterior object to said vehicle, said measuring means comprising radar” The
`
`Board defined this as radar.
`
`32.
`
`“dimming the headlights” The Board defined this as decreasing the
`
`intensity or output of the headlight to a lower level of illumination.
`
`11
`
`
`
`33.
`
`“wherein said categories further comprise radiation from taillights of
`
`a vehicle-in-front” The Board defined this as covering categorizing radiation from
`
`taillights of a vehicle-in-front, which may include additional types of radiation.
`
`C.
`
`34.
`
`Person of Ordinary Skill in the Art
`
`I understand that all my opinions with respect to the validity
`
`(including claim construction) of the ‘000 patent are to be considered from the
`
`viewpoint of the hypothetical person of ordinary skill in the art as of the date of the
`
`invention. I understand that this hypothetical person of ordinary skill in the art is
`
`considered to have the normal skills and knowledge of a person in a certain
`
`technical field, as of the time of the invention at issue. I understand that the factors
`
`that may be considered in determining the level of ordinary skill in the art include
`
`the education level of the inventor, the types of problems encountered in the art,
`
`prior art solutions to those problems, the educational level of active workers in the
`
`field, the rapidity with which innovations are made, and the sophistication of the
`
`technology.
`
`35.
`
`In my opinion, based on my experience and knowledge in the field,
`
`such a person would have at least a bachelor’s degree in a relevant engineering
`
`field and at least some professional experience, perhaps two to three years,
`
`working with exterior monitoring or object detecting systems as well as pattern
`
`12
`
`
`
`recognition methods, or such a person can have more experience or education such
`
`as a master’s degree or doctorate degree.
`
`36.
`
`I have read the opinion of Dr. Papanikolopoulos, and I have no
`
`fundamental dispute with his proposed definition. Therefore, I have no objection
`
`to using it for the purposes of my analysis.
`
`D.
`
`37.
`
`Scope and Content of the Prior Art
`
`In my opinion, the scope and content of the prior art would have been
`
`narrower than that offered by Dr. Papanikolopoulos. In my opinion, the scope and
`
`content of the prior art would not have generically included any “vehicle sensing
`
`systems,” as there are many vehicle sensing systems that are not specifically
`
`configured to identify of an exterior object or a source of radiation. In my opinion,
`
`the scope and content of the prior art that is relevant here, is that which would have
`
`included sensors and pattern recognition algorithms for object or radiation source
`
`identification, including those for automotive use.
`
`38.
`
`I do not disagree that the references offered by Dr. Papanikolopoulos
`
`and applied by the Board in its Institution Decision are within the scope and
`
`content of the prior art to the ‘000 patent. I do, however, disagree that any of the
`
`references invalidate the ‘000 patent claims that I address below.
`
`V.
`
`LEGAL STANDARDS APPLIED
`
`13
`
`
`
`39.
`
`I am not an expert in patent law, and I am not purporting to provide
`
`any opinions regarding the correct legal standards to apply in these proceedings. I
`
`have been asked, however, to provide my opinions in the context of the following
`
`legal standards that have been provided to me by AVS’s attorneys.
`
`40.
`
`Anticipation: It is my understanding that a patent is invalid as
`
`anticipated if each and every limitation of the claimed invention is disclosed in a
`
`single prior art reference, either expressly or inherently, such that one of ordinary
`
`skill in the art would be enabled to make the claimed invention without undue
`
`experimentation. For anticipation, every limitation of a claim must appear in a
`
`single prior art reference as arranged in the claim. An anticipating reference must
`
`describe the patented subject matter with clarity and detail to establish that the
`
`subject matter existed in the prior art and that such existence would be recognized
`
`by one of ordinary skill. The prior art is enabling if the disclosure would have put
`
`the public in possession (i.e., provided knowledge) of the claimed invention and
`
`would have enabled one of ordinary skill to make or carry out the invention
`
`without undue experimentation.
`
`41.
`
`Inherency: I understand that if a prior art reference does not expressly
`
`disclose a claimed feature, but the teaching of the reference would necessarily
`
`result in a product with the claimed feature, then anticipation may be met
`
`inherently. For a prior art reference to inherently disclose a claimed feature,
`
`14
`
`
`
`however, the feature must be necessarily present and may not be established just
`
`because it may be probable or possible. The mere fact that a condition may result
`
`from a set of circumstances, or even probably results from the set of circumstances,
`
`is not sufficient for proof of inherency. Further, I understand that for the purposes
`
`of evaluating anticipation of a prior art reference, the reference must be interpreted
`
`from the understanding of one of ordinary skill in the art.
`
`42. Obviousness in General: I have been informed that a patent can also
`
`be invalidated through obviousness if the subject matter of a claim as a whole
`
`would have been obvious at the time of the invention to a person of ordinary skill
`
`in the art. I understand that obviousness allows for the combination of prior art
`
`references. I have been informed that there are three basic inquiries that must be
`
`considered for obviousness:
`
`a. What is the scope and content of the prior art?
`
`b. What are the differences, if any, between the prior art and each claim
`
`of the patent?
`
`c. What is the level of ordinary skill in the art at the time the invention
`
`of the patent was made?
`
`I also understand that when prior art references require selective combination to
`
`render a patent obvious, there must be some reason to combine the references other
`
`than hindsight. Even if there would have been an apparent reason for combining
`
`15
`
`
`
`prior art references, however, there must also have been a reasonable expectation
`
`of success. I understand that features from prior art references need not be
`
`physically combinable (i.e., a combination may be obvious if one of ordinary skill
`
`in the art would know how to make any necessary modifications to combine
`
`features from prior art references), but that this concept does not negate the
`
`requirement of a reasonable expectation of success. One must also consider the
`
`evidence from secondary considerations including commercial success, copying,
`
`long-felt but unresolved needs, failure of others to solve the problem, unexpected
`
`results, and whether the invention was made independently by others at the same
`
`time of the invention.
`
`I understand that these secondary considerations can
`
`overcome a finding of obviousness.
`
`VI. OPINIONS REGARDING VALIDITY OF ‘000 PATENT CLAIMS
`
`A.
`
`None of the Cited References Specifically Disclose a Pattern
`Recognition Algorithm Generated From Data of Possible Exterior
`Objects and Patterns of Received Electromagnetic Illumination
`from the Possible Exterior Objects (claims 10, 11, 19, and 23) or a
`Pattern Recognition Algorithm Generated from Data of Possible
`Sources of Radiation and Patterns of Received Radiation from the
`Possible Sources (claims 16, 17, and 20)
`
`43. As I previously discussed, independent claims 10 and 16 both require
`
`a “trained pattern recognition means.” In claim 10, the “trained pattern recognition
`
`means” is “structured and arranged to apply a pattern recognition algorithm
`
`generated from data of possible exterior objects and patterns of received
`
`16
`
`
`
`electromagnetic illumination from the possible exterior objects.” (‘000 patent at
`
`claim 10.) Therefore, dependent claims 11 and 19 also require this limitation. In
`
`claim 16, the “trained pattern recognition means” is “structured and arranged to
`
`apply a pattern recognition algorithm generated from data of possible sources of
`
`radiation including lights of vehicles and patterns of received radiation from the
`
`possible sources.” (‘000 patent at claim 16.) Therefore, dependent claims 17 and
`
`20 also require this limitation. In independent claim 23, the method includes
`
`“generating a pattern recognition algorithm from data of possible exterior objects
`
`and patterns of received electromagnetic illumination from the possible exterior
`
`objects.”
`
`44.
`
`The only prior art reference, according to the Board’s preliminary
`
`Decision to Institute Inter Partes Review, which allegedly discloses a “trained
`
`pattern recognition means” comprising “a pattern recognition algorithm generated
`
`from data of possible exterior objects and patterns of received electromagnetic
`
`illumination from the possible exterior objects” (claims 10, 11 and 19); a “trained
`
`pattern recognition means” comprising “a pattern recognition algorithm generated
`
`from data of possible sources of radiation … and patterns of received radiation
`
`from the possible sources” (claims 16, 17, and 20); and “generating a pattern
`
`recognition algorithm from data of possible exterior objects and patterns of
`
`received electromagnetic illumination from the possible exterior objects” (claim
`
`17
`
`
`
`23), is Lemelson. (See Paper 16, Board Decision at 26-40 and 44.) Therefore, my
`
`understanding is that the prior art that is pertinent to this review (validity of these
`
`claims) is that disclosed by Lemelson and thus my considerations should be
`
`restricted to the disclosures of Lemelson as to the subject limitation.
`
`In other
`
`words, the Board did not allege that any other reference rendered obvious: “trained
`
`pattern recognition means” comprising “a pattern recognition algorithm generated
`
`from data of possible exterior objects and patterns of received electromagnetic
`
`illumination from the possible exterior objects;” a “trained pattern recognition
`
`means” comprising “a pattern recognition algorithm generated from data of
`
`possible sources of radiation … and patterns of received radiation from the
`
`possible sources;” and “generating a pattern recognition algorithm from data of
`
`possible exterior objects and patterns of received electromagnetic illumination
`
`from the possible exterior objects.” I understand that the Board’s decision to
`
`Institute Inter Partes Review is final as to grounds not adopted. (See Paper 16,
`
`Board Decision at 45 (“FURTHER ORDERED that all other grounds raised in
`
`Toyota’s petition are denied.”).)
`
`45. Accordingly, I understand that if Lemelson is found to not disclose or
`
`teach “trained pattern recognition means” comprising “a pattern recognition
`
`algorithm generated from data of possible exterior objects and patterns of received
`
`electromagnetic illumination from the possible exterior objects;” “trained pattern
`
`18
`
`
`
`recognition means” comprising “a pattern recognition algorithm generated from
`
`data of possible sources of radiation … and patterns of received radiation from the
`
`possible sources;” and “generating a pattern recognition algorithm from data of
`
`possible exterior objects and patterns of received electromagnetic illumination
`
`from the possible exterior objects,” then claims 10, 11, 16, 17, 19, 20, and 23
`
`overcome all remaining grounds and will be upheld in the inter partes review. For
`
`the reasons discussed below, in my opinion, Lemelson does not disclose, either
`
`expressly or inherently, or teach the required “trained pattern recognition means”
`
`comprising “a pattern recognition algorithm generated from data of possible
`
`exterior objects and patterns of received electromagnetic illumination from the
`
`possible exterior objects;” “trained pattern recognition means” comprising “a
`
`pattern recognition algorithm generated from data of possible sources of radiation
`
`… and patterns of received radiation from the possible sources;” and “generating a
`
`pattern recognition algorithm from data of possible exterior objects and patterns of
`
`received electromagnetic illumination from the possible exterior objects.” For at
`
`least
`
`the reasons described below, and considering
`
`the above-mentioned
`
`constraints, in my opinion, the prior art does not render claims 10, 11, 16, 17, 19,
`
`20, and 23 unpatentable.
`
`a) Lemelson
`
`19
`
`
`
`Lemelson’s Disclosure of Training a Pattern Recognition
`Algorithm—Lemelson Does Not Expressly Disclose the Type and
`Nature of the Training or “Known Inputs” Used for Training
`
`46.
`
`Lemelson discloses a system for identifying objects exterior to a
`
`vehicle, and it does disclose using a type of pattern recognition algorithm, a neural
`
`network. (See Exhibit 1002, Lemelson at Fig 3, 7:47, 8:1) The only pertinent
`
`discussion in Lemelson, however, relating to generating the neural network, is
`
`found at column 8, line 4, which states “[t]raining involves providing known inputs
`
`to the network resulting in desired output responses.” (Id. at 8:4-6.)
`
`47.
`
`I note that this is also the only reference to training in Lemelson found
`
`in the declaration of Dr. Papanikolopoulos. (See Exhibit 1013, Papanikolopoulos
`
`Decl. at ¶¶ 47-65.) Specifically, Dr. Papanikolopoulos quotes the sentence from
`
`Lemelson that states “[t]raining involves providing known inputs to the network
`
`resulting in desired output responses.” (Id. at ¶59.) Nowhere else in his
`
`declaration does he use the words “training”, “trained,” or “generate” when
`
`discussing the Lemelson reference.
`
`Dr. Papanikolopoulos’s Deposition Testimony
`
`48. At his deposition, Dr. Papanikolopoulos tried to suggest that other
`
`disclosure in Lemelson also allegedly relates to the nature and type of training
`
`involved. (See Exhibit 2003, Papanikolopoulos Dep. Tr. at 163:4-165:13, 167:4-
`
`169:22.) But in fact, the disclosure pointed to by Dr. Papanikolopoulos points out
`
`20
`
`
`
`a specific neural network structure and where it is integrated into Lemelson's
`
`description, but it does not relate to the training per se of a pattern recognition
`
`algorithm. For example, Dr. Papanikolopoulos referred to Figures 1-5 at his
`
`deposition. None of those Figures refers to training or includes the word training.
`
`Each of those figures only relates to how the system gathers data while in use
`
`(after it has been trained). Dr. Papanikolopoulos’s argument, for example, that
`
`Figure 1’s reference to an “image analyzing computer” discloses the nature and
`
`extent of the “training” is unfounded. (See Exhibit 1002, Lemelson at Fig. 1.) The
`
`fact that a computer was used to analyze images obtained when the system is in use
`
`(after it has been trained) tells nothing about the nature and extent of the training
`
`phase. An image analysis computer could be used to analyze received camera
`
`images when the vehicle is driving along the road, but the system could have been
`
`trained with something else entirely (such as simulated inputs as I discuss below).
`
`Nor does the fact that Figure 5 refers to “images” provides any information at all
`
`about the training phase of the algorithm. Again, the “images” referred to in
`
`Figure 5 are the images gathered by the camera/receiver when the vehicle is being
`
`driven in normal operating mode (post-training), to identify objects—nothing in
`
`that Figure refers to the training phase of the algorithm. (See id. at Fig. 5.)
`
`49. Dr. Papanikolopoulos appears to conjecture that the way a pattern
`
`recognition system is used after training necessarily discloses the methods and
`
`21
`
`
`
`means by which it was trained during the training phase, and this is where we
`
`sharply differ. There are many different modes in which the training phase can be
`
`conducted independently of the precise intended use in the actual operating phase
`
`of the application. In this case for example, the training phase can be conducted in
`
`real time while driving around a prototype in real conditions, or it can be
`
`conducted by storing sensor data from driving a prototype in the actual
`
`environment and later using these data in a lab with or without pre-processing, or it
`
`can be conducted in the controlled environment of a lab in which real-life
`
`conditions are re-created, or it can be conducted with completely simulated data
`
`generated by software in a lab, etc. My point is that the intended u