HW1- Reading Assignment on Intelligent SystemsIssued: 3 Jan ’18 Due:17 Jan ’18Instructions· Submission should be typed with answersin a different font color.
· Submitassignment as a MSWord or PDF file to the ECE/SYS6410 web page on http://moodle.oakland.edu. · Labelyour file with an identifiable name such as CheokKaC_HW1.
docx.Purpose· Expandyour horizon on the disciplines and context of artificial intelligence (AI) andvarious disciplines that follow: computational intelligence, machineintelligence, intelligent controls, etc.· Expandyour thoughts on the AI topics are applied to control systems. · Tobecome fascinated at the potential of future AI and RoboticsResources· For general knowledge, search theinternet (Google or other search engines) with keywords.
While o Wikipedia is generally a good start.But there are many other sites that may have better info.o Read up articles, magazines, or bookson the subject· Forrigorous technical/math detailo Search http://scholar.google.com/ o Being an OU student, you have theprivilege to access a huge collection of scientific & technical articles http://library.oakland.edu/articles/ejournal.htm.
Take advantage of receiving free prints asappropriate!Goals· You are to enjoy discovering the worldof AI along with rapid advances of today’s technologies. · Forma clearer concept, description, methods, application, highlights, potential,fascination, complexity and future of the subject matters.· Youmay find that it is complicated to encapsulate the subject matter, i.e., thereis no one simple concept.· Answerthe questions below. They are generallysubjective and meant to stimulate thoughts. Task1: Artificial intelligent systems (50%)Start with https://en.
wikipedia.org/wiki/Artificial_intelligence. Read the article and look up its referencelinks as necessary. Then add one to twosentences for each of categories regarding what it can do or what it is. Problem a. provided a partial sample.
1. Reasoning,problem solving: Reasoning and problem solving are important problems of AI.Earlier AI algorithms that researchers developed were imitating thestep-by-step reasoning that humans use to solve problems or make decisions.
These algorithms were intended to solve puzzles, games or economic problems.2. Knowledgerepresentation:Knowledge representation is the representation ofinformation of world such as objects, their properties, categories and relationbetween them; situations and events; causes and effects; etc., in a form inwhich a computer can utilize to solve complex tasks.Knowledge representation aids Artificial Intelligence (AI)to solve complex tasks like diagnosis of medical condition of patients, havingdialogues in natural language, knowledge based searchand services (like Yebol), etc.3.
Planning:AI planning involves realization of strategies that are tobe executed by intelligent agents, which help them to visualize the future andmake predictions on how their actions would change it. Unlike classicalplanning approach, in AI planning the agents do not assume that it is not theonly actor and adopt based on the actions of other agents.AI planning helps autonomous robots, autonomous groundvehicles and unmanned aerial vehicles to plan their trajectories of motion andthe actions it needs to take to maximize the chances of achieving the desiredtrajectories.4. Learning:Learning is the process of acquiring new knowledge ormodifying the existing knowledge. Machine learning facilitates computers theability to learn and improve automatically without being explicitly programed.Machine learning finds application in optical character andspeech recognition, game playing, autonomous vehicles, financial marketanalysis, internet fraud detection, machine learning control systems, etc.5.
Naturallanguage processing:Natural language processing facilitates intelligent agentsthe ability to understand human language. Speech recognition, languagetranslation, analysis of text, language parsing and understanding, questionanswering, etc.Natural language processing find application in machinetranslation, question answering, information retrieval and text mining.6. Perception:Machine perception is a form of AI that provides thecomputer the ability to interpret data from the sensors, in a way that issimilar to the human perception of the world around them using their senses.
Machine perception have applications in speech recognition,image recognition and object recognition.7. Motionand manipulationMotion planning and object manipulation are subfields of AIand are of immense importance to robotics. They help a robot or otherintelligent agent to learn about its surroundings and decide on how to get fromone point to another.8. Socialintelligence:Social intelligence of AInetworks is an important topic because it helps the AI agents to take betterdecisions by predicting the actions of other humans by reading andunderstanding their motives and emotional state.In game theory, socialintelligence is vital as the AI agent must be able to detect the emotionaldynamics of the human interaction and take decisions accordingly to improve theoverall experience.9.
Creativity:Artificial creativity orcomputational creativity is a sub field of AI that deals to replicate humanlevel creativity using a computer.Artificial creativity findsuses in computer story generation, musical creativity, etc.10. Generalintelligence:Artificial general intelligence is anintelligence that helps a machine to perform intellectual tasks that a humancan. 11. Cyberneticsand brain simulation:Cybernetics and brainsimulation is an approach of AI that creates mathematical models and performtheoretical analysis of brain to understand the principles behind itsfunctioning, its structure and development, how information is passed etc.12. Symbolic:Symbolic artificialintelligence is an approach of AI that is based on high level symbolicrepresentation of problems, logic and search.
This approach assumes that many aspects of intelligence can be achieved by themanipulation of symbols. The nature of symbolic AI is to acceptsymbolic input information and create symbolic output information.Symbolic AI approach can befound in computer expert systems (an expert computer emulates thedecision-making ability of a human expert.) which use a network that connectssymbols and rules.
The expert system processes these rules to take decisionsand determines the additional information it needs.This method is easier to easier to explain, easier todebug and better for abstract problems.13. Sub-symbolic:Sub-symbolicAI approaches intelligence without specific representations of knowledge. In this approach, information is parallellyprocessed by mathematical calculations (example neural networks, Bayesianlearning, deep learning etc.).
Theseapproaches are more robust against noise, require less knowledge upfront,easier to scale and is more useful for connecting to neuroscience14. Statistical:Statistical approaches orevidence based approaches in AI involve using sophisticated mathematical toolsto solve specific sub problems. It uses optimization techniques toautomatically improve the performance of a piece of software, based on evidencepresent in measurement data.
Statistical approaches canhandle uncertainties better than the symbolic and sub-symbolic approaches.15. Integratingthe approaches:Some simple AI agents whichsolve a specific problem can use any approach which works, i.e., symbolic,sub-symbolic or statistical or any new approaches. Compilated agents like anautonomous robot network, autonomous vehicles etc., will often have to takeapproaches that integrate all these methods to be able to perform the complextasks that they would be given.
16. Search and optimization:Search algorithms helps tosolve many problems in AI by searching through viable solutions. For real worldproblems, simple search algorithms are rarely sufficient as they are either tooslow or never completes. The solution for real world problems is to use searchalgorithms based on optimization techniques. Search algorithms retrieveinformation stored within some data structure and the optimization techniquesallow the computer to select the best element, based on some criterion, fromset of available alternatives. This method can be thought of beginning withsome form of a guess (random or intelligent) and the continuously refining theguess an incremental basis, until there is no further scope for refinement.
Optimization based searchtechnique can be found in path-planning of autonomous vehicles or robots, etc.17. Logic:Logical AI uses set of statements that can betrue or false along with quantifiers and predicates, and express facts aboutobjects, their properties, and their relations with each other. Fuzzy logic isan example of logical AI.
Fuzzy logic based control systems can be found inmodern consumer products.18. Probabilistic methods foruncertain reasoning:Probabilistic methods in AI employ probability theory andeconomics to solve some problems in AI that have incomplete or uncertaininformation. Some of the tools that are developed using these methods areBayesian networks, hidden Markov models, Kalman filter, particle filter, etc.Bayesian networks find applications in image processing,data fusion, risk analysis, document re classification, bioinformatics, etc.Hidden Markov models find applications in computational finance, machinetranslation, single molecule kinetic analysis, etc.
Kalman filters findapplications in inertial guidance systems, nuclear medicine, weatherforecasting, etc. Particle filters find application in image processing,molecular chemistry, computational physics etc.19. Classifiersand statistical learning methods:Statisticalclassification and learning methods employ functions that use patternmatching to determine aclosest matching category for the new data based on previous experience. Statistical learning theory deals with the problem offinding a predictive function based on data. These learning methods havebecome a powerful weapon to overcome uncertainty in AI scenarios and,consequently, it has been widely implemented in many modern AI frameworks.Applications include datamining, e-mail spam filtering etc.
20. Neural networks:Artificial Neural Network’s (ANN) are inspired by biologicalneural networks and consists of interconnected group of neurons, where eachneuron is represented by a mathematical model. A trained ANN determines outputresponse to the input signal depending on the training values and themathematical functions of the neurons. Main categories of neural network arefeedforward neural networks and recurrent neural networks (allows feedback andshort-term memories of previous inputs).ANN find applications in system identification, controltheory, pattern recognition, automated trading systems, quantum chemistry etc.21.
Deep feedforward neural networks:Deep feedforward neural network consists of substantialnumber of layers of neurons arranged in a feedforward way. They may contain many layers of non-linear hiddenunits and a very large output layer. These networks are trained onelayer at a time.
The recent growth in computational power have resulted indevelopment of more efficient methods of training of deep networks.Deep feedforward neuralnetworks find applications in automatic speech recognition, bio informatics,mobile advertising, image restoration, autonomous driving, etc.22. Deep recurrent neural networks:Deeprecurrent neural networks have many layers of interconnected units ofartificial neural networks and the connections between units form a cycle. Theyconstitute to deep learning ANN algorithms in which learning is based on datarepresentations, as opposed to task-specific.Deeprecurrent neural applications are applied by Google, Microsoft and Baidu toimprove the performance of speech recognition. 23. Languages:There are several languages that are developed specially forprogramming of artificial intelligence algorithms such as Python, Prolog, Lisp,etc.
Apart from these, there are also several other general-purpose programminglanguages like C++ and simulation tools like MATLAB and SIMULINK to developartificial intelligence algorithms.24. Control theory:Intelligent Control is a branch of control theory that usevarious artificial intelligence computational approaches like neural networks,fuzzy, neuro-fuzzy control, Bayesian control etc.Intelligent control finds applications in flight controlsystems, autonomous driving systems, oil drilling processes etc.25. Competitionand prizes:To promote research anddevelopment in AI, there have been established several competitions and prizes.
The principal areas where there are general machine intelligence,conversational behavior, data-mining, robotic cars, robot soccer and games.26. Healthcare:AI in healthcare uses algorithms to analyze relationships between prevention or treatmenttechniques and patient outcomes by analysis of complex medical data.Usage of AI appeared several times recently, for example,Microsoft’s system to assist doctors to find the right treatment of cancer,IBM’s Watson which not only won a game show but also successfully diagnosed awoman suffering from leukemia, autonomous robot that performed surgery, etc.27. Automotive:AI facilitates the creation and evolution of self-drivingcars and trucks.
As of 2016, reports suggest that there are more than 30companies that are utilizing AI in the creation of driverless cars, to namesome companies, they are Tesla, Google and Apple. These vehicles incorporatecomplex AI algorithms, along with complex control strategies to performfunctions such as autonomousbraking, autonomous lane changing, collision prevention, navigation, mapping,trajectory planning, etc. based on the available data.
28. Finance and economics:In finance and economic sector AI is used to prevent fraud protectionby detecting abnormal behavioral patterns of users. AI is also used in bookkeeping, stock investment, online trading, estimation of supply demand curves,reduction of information asymmetry.29. Video games:AI is also used in video games to generate adaptive andhuman like intelligent behaviors from non-playing characters. It finds usage incomputer board games like chess, go, checkers, poker players, scrabble, etc.,and many other types of games30. Education in AI:In the recent times many private bootcamps, free programsand paid programs have been developed to facilitate education in AI.
A recentstudy concluded that there is a shortage of about 1.5 skilled AI professionals.These bootcamps have developed some programs to meet that demand like The DataIncubator and General Assembly.31. Partnership on AI:Companies like Amazon, Google, Facebook, Microsoft and IBMhave recently established a non-profit partnership whose aim is to conductresearch, share best practices, create educational material, consult therelevant third parties, respond toquestions from the public and media, etc. Partnerships like these will help advance thefield of AI by formulating itsbest practices and serving as a platform about artificial intelligence.32. AlanTuring’s “polite convention”:Alan Turing developed atest in 1950, called the Turing test, which is a test of a machine’s ability toexhibit intelligent behavior equivalent to that of a human.
Turing proposedthat a human evaluator would judge natural language conversations between ahuman and a machine that is designed to generate human-like responses.In the Turing test, a humanwould evaluate natural language conversations between a human and a machine andthe machine is said to have passed the test if the human evaluator cannotreliably distinguish between the responses of the computer and the responses ofthe human.33. TheAI Effect:In the field of AI, atechnology is no longer considered as AI and is removed from the list if itreaches mainstream use. This phenomenon is called AI effect. Task 2: Deep learning as an essence of IntelligentControl System (50%)Doa Google Scholar search on “self-driving car” and select the link called”end-to-end learning for self-driving cars,” by M Borjaski, et al.
It should take you to Cornell UniversityLibrary page, displaying the titled paper. Download the pdf. Or here’s thedirect link https://arxiv.
org/pdf/1604.07316.pdf.Readthe paper as best as you can, without the a priori understanding of thetechnique they used. It’s well writtenwith clear explanation if the experimental process, data acquisition, deeplearning training of convolutional neural network, successful on-roadverification, and detail explanation of how CNN layers tries to learn. Thefollowing video is complementary to the above paper.
Enjoy.Published on Apr 6,2016NVIDIA CEO Jen-Hsun Huang describes howthe open platform NVIDIA DRIVE uses AI to advance the driving experience, frominfotainment to autonomous vehicles, and how the data gathered fromsensor-filled cars will be used to create comprehensive HD maps, at the GPUTechnology Conference.https://www.youtube.com/watch?v=RVmV9SXJeBg&feature=youtu.
beAfterunderstanding the paper and video, explain in a simple way how this is similarto how a new driver learns to drive a car. In other words, compare the above experiments to teaching a kid how todrive. A paragraph of several sentences.Answer:Ingeneral, the process of a human driving an automobile consists of three steps,sensing, planning and acting, i.e.
, sense and perceive the environment (roadconditions, temperature, etc.), make decisions in response to thoseperceptions, and control the actuators (throttle, brakes, steering, gear, etc.)to achieve the desired motion. This philosophy also applies to autonomousdriving. The reference paper proposes an end to end AI strategy using neuralnetwork and deep learning for autonomous driving, with emphasis on the methodof its training.
Thestrategy proposed was to sense the environment using 3 cameras and feed it to adeep CNN. The CNN would calculate the desired trajectory and propose a steeringangle command to achieve this trajectory. This proposed steering wheel commandwould then be compared to the reference steering wheel angle generated by thehuman driver. The error between these two signals would be used to train theCNN and the weights are adjusted in a direction which reduces this error. Themethod of training the CNN that is similar to how we teach a new driver todrive the car. A student driver drives the car in the supervision of anexperienced driver who sits in the passenger seat. The student driver sensesthe environment and applies the control commands accordingly, whilecontinuously taking inputs from the experienced driver and learning from themistakes.
Initiallystudent driver starts by driving simple maneuvers (for example lane following)in simple road conditions (for example dry roads with clear lane markings,etc.) and simple traffic conditions and incrementally learns by supervision onhow to handle complex maneuvers and uncertain scenarios. This is similar to theapproach in the paper, where the neural network was first fed with simple data,for example the data where driver was just staying in the lane. Data consistingof more complex scenarios, for example roads without lane markings, residentialroads, parked cars, unpaved roads, tunnels, foggy conditions, low frictionsurfaces etc. were incrementally fed to the CNN and made it learn.