SOLUTION: Artificial Intelligence | My Assignment Tutor

Artificial Intelligence (AI)AI is the futureArtificial Intelligence technologies are fastevolving which are widely adopted intechnology & business circles.Many governments and industries see thatAI is the future; however, if we lookaround us and check our daily ilfeactivities, AI does not like it is the future;it is the present, and the future will bemore advanced and sophisticated.So What is AI?AI technology is the branch of computerscience that accentuates the building ofintelligent machnines that will work andthink like humans. At first, the Facebookapplication has adopted AI for imagerecognition. Other AI example is speechrecognition which uses artificial neuralnetworks.With the advancement of technology, we are connected tovarious types of artificial intelligence products, whether it isSiri, Alexa, or Watson.What is Machine Learning?“Learning is any process by which systemimproves performance from experience” –Herbert SimonMachine Learning is the study of algorithmsthat:– Improves their performance P– At some task T– With experience E.A well-defined learning task is given by ComputerDataProgramOutputComputerDataOutputProgramTraditional LearningMachine LearningCharacterisation of Intelligent SystemProcess one or more of these:– Capability to extract and store knowledge– Human like reasoning process– Learning from experience (or training)– Dealing with imprecise experience of facts– Finding solutions through processes similar to naturalevolution.Characterisation of Intelligent System (cont.)More sophisticated Interaction with the user through:– Natural language understanding– Speech recognition and synthesis– Image analysisExpert SystemsExpert systems are designed to solve complex problems byreasoning through bodies of knowledge, represented mainly as ifthen rules rather through conventional procedural code.Stochastic SystemsA stochastic system is a system whose future states, due to itscomponents’ possible interactions, are not known preciselyAlgorithms and TechniquesModelling Algorithms:– Prediction, classification, pattern recognition, time seriesprediction, image recognition, speech recognition, etc.Search Algorithms:– Parameters search, optimisation, path finding, scheduling,system tuning, etc.Modelling AlgorithmsWhat are Neural Networks?• Models of the brain and nervous system• Highly parallel– Process information much more like the brain than a serial computer• Learning• Very simple principles• Very complex behaviours• Applications– As powerful problem solvers– As biological modelsNeural Networks HistoryHow does a simple neural network work?Information flows through a neural network in two ways.When it’s learning (being trained) or operating normally(after being trained), patterns of information are fed into thenetwork via the input units, which trigger the layers ofhidden units, and these in turn arrive at the output unitsANNs – The basics• ANNs incorporate the two fundamental components ofbiological neural nets:1. Neurones (nodes)2. Synapses (weights)Neurone vs. NodeStructure of a node:Squashing functionlimits node output:Feed-forward nets• Information flow is unidirectional• Data is presented to Input layer• Passed on to Hidden Layer• Passed on to Output layer• Information is distributed• Information processing is parallelInternal representation(interpretation) of dataFeeding data through the net:0.3775110.5=+ eSquashing:(1  0.25) + (0.5  (-1.5)) = 0.25 + (-0.75) = – 0.5• Data is presented to the network in the form of activations inthe input layer• Examples– Pixel intensity (for pictures)– Molecule concentrations (for artificial nose)– Share prices (for stock market prediction)• Data usually requires pre-processing– Analogous to senses in biology• How to represent more abstract data, e.g. a name?– Choose a pattern, e.g.0-0-1 for “Chris” and 0-1-0 for “Becky”Weight settings determine the behaviour of a network.How can we find the right weights?Training the Network – Learning• Backpropagation– Requires training set (input / output pairs)– Starts with small random weights– Error is used to adjust weights (supervised learning)→ Gradient descent on error landscape • Advantages– It works!– Relatively fast• Downsides– Requires a training set– Can be slow– Probably not biologically realistic• Alternatives to Backpropagation (Hebbian learning;Reinforcement learning; Artificial evolution)Biological Neural Nets• Pigeons as art experts (Watanabe et al. 1995)Experiment:• Pigeon in Skinner box (the pigeons were trained bySkinner to peck at a target, and they rewarded with foodwhen they completed the task correctly)• Present paintings of two different artists (e.g. Chagall /Van Gogh)• Reward for pecking when presented a particular artist (e.g.Van Gogh)Marc ChagallVan GoghWhich one for Chagall and which one for Van Gogh?What about these two?• Pigeons were able to discriminate between Van Gogh andChagall with 95% accuracy (when presented with picturesthey had been trained on)• Discrimination still 85% successful for previously unseenpaintings of the artists• Pigeons do not simply memorise the pictures• They can extract and recognise patterns (the ‘style’)• They generalise from the already seen to make predictions• This is what neural networks (biological andartificial) are good at (unlike conventional computer)Example: Voice Recognition• Task: Learn to discriminate between two differentvoices saying “Hello”• Data– Sources• Voice A:• Voice B:– Format• Frequency distribution (60 bins)• Analogy: cochleaNetwork architecture– Feed forward network• 60 input (one for each frequency bin)• 6 hidden• 2 output (0-1 for “voice A”, 1-0 for “voice B”)Presenting the dataVoice AVoice BPresenting the data (untrained network)0.430.260.730.55Voice AVoice BCalculate error0.43 – 0 = 0.430.26 –1 = 0.740.73 – 1 = 0.270.55 – 0 = 0.55Voice AVoice BBackprop error and adjust weights0.43 – 0 = 0.430.26 – 1 = 0.740.73 – 1 = 0.270.55 – 0 = 0.551.170.82Voice AVoice B• Repeat process (sweep) for all training pairs– Present data– Calculate error– Backpropagate error– Adjust weights• Repeat process multiple timesPresenting the data (trained network)0.010.990.990.01Voice AVoice BResults – Voice Recognition– Performance of trained network• Discrimination accuracy between known “Hello”s:100%• Discrimination accuracy between new “Hello”’s:100%Results – Voice Recognition (cont.)– Network has learnt to generalise from original data– Networks with different weight settings can havesame functionality– Trained networks ‘concentrate’ on lowerfrequencies– Network is robust against non-functioning nodesDeep Learning– Deep learning is an AI function that mimics theworkings of the human brain in processing data for usein detecting objects, recognizing speech, translatinglanguages, and making decisions.– Deep learning AI is able to learn without humansupervision, drawing from data that is bothunstructured and unlabelled.– Deep learning, a form of machine learning, can be usedto help detect fraud or money laundering, among otherfunctions.How Deep Learning Works– Deep learning unravels huge amounts of unstructured data thatwould normally take humans decades to understand andprocess.– A traditional approach to detecting fraud or money launderingmight rely on the amount of transaction that ensues, while adeep learning nonlinear technique would include time,geographic location, IP address, type of retailer, and any otherfeature that is likely to point to fraudulent activity.How Deep Learning Works– The first layer of the neural network processes a raw datainput like the amount of the transaction and passes it on to thenext layer as output. The second layer processes the previouslayer’s information by including additional information likethe user’s IP address and passes on its result.– The next layer takes the second layer’s information andincludes raw data like geographic location and makes themachine’s pattern even better. This continues across all levelsof the neuron network.Fuzzy LogicFuzzy logic is an approach to computing based on“degrees of truth” rather than the usual “true or false”(1 or 0) Boolean logic on which the modern computeris based.Fuzzy LogicFuzzy logic seems closer tothe way our brains work. Weaggregate data and form anumber of partial truthswhich we aggregate furtherinto higher truths which inturn, when certain thresholdsare exceeded, cause certainfurther results such as motorreaction.How Does Fuzzy Logic Work?Fuzzy logic systems architecture has four main parts:Fuzzification Module – It transforms the system inputs,which are crisp numbers, into fuzzy sets. It splits theinput signal into five steps such as: LPx is Large PositiveMPx is Medium PositiveSx is SmallMNx is Medium NegativeLNx is Large Negative Knowledge Base – It stores IF-THEN rules provided byexpert.Inference Engine – It simulates the human reasoningprocess by making fuzzy inference on the inputs and IFTHEN rules.Defuzzification Module – It transforms the fuzzy setobtained by the inference engine into crisp value.Crisp InputFuzzifierRulesDefuzzifierIntelligenceCrisp OutputFuzzyinput setFuzzyoutput setFuzzy logic systems architectureMembership FunctionMembership functions allow you to quantify linguisticterm and represent a fuzzy set graphically.1 00.5Membership functionLN MN S MP LP-10 -5 0 5 10Input voltageExample of a Fuzzy Logic SystemLet us consider an air conditioning system with 5-levelfuzzy logic system. This system adjusts the temperatureof air conditioner by comparing the room temperatureand the target temperature valueRoomtemperatureAir ConditionerFuzzy LogicSystemCommand:HeatCoolNo changeTargetTemperatureAlgorithm– Define linguistic Variables and terms (start)– Construct membership functions for them. (start)– Construct knowledge base of rules (start)– Convert crisp data into fuzzy data sets using membershipfunctions. (fuzzification)– Evaluate rules in the rule base. (Inference Engine)– Combine results from each rule. (Inference Engine)– Convert output data into non-fuzzy values.(defuzzification)Step 1: Define linguistic Variables and terms– Linguistic variables are input and output variables in theform of simple words or sentences. For room temperature,cold, warm, hot, etc., are linguistic terms.– Temperature (t) = {very-cold, cold, warm, very-warm,hot}– Every member of this set is a linguistic term and it cancover some portion of overall temperature values.Step 2: Construct membership functions for themThe membership functions of temperature variable are:1 00.5Membership functionLN MN S MP LP0 10 20 30 40 Input temperatureStep 3: Construct knowledge base rulesCreate a matrix of room temperature values versus targettemperature values that an air conditioning system isexpected to provide. Room Temp /TargetVery_ColdColdWarmHotVery_HotVery_ColdNo_ChangeHeatHeatHeatHeatColdCoolNo_ChangeHeatHeatHeatWarmCoolCoolNo_ChangeHeatHeatHotCoolCoolCoolNo_ChangeHeatVery_HotCoolCoolCoolCoolNo_Change Step 3: Construct knowledge base rules (cont.)Build a set of rules into the knowledge base in the form ofIF-THEN-ELSE structures. ConditionAction1IF temperature = (Cold OR Very_Cold) AND target = Warm THENHeat2IF temperature = (Hot OR Very_Hot) AND target = Warm THENCool3IF (temperature = Warm) AND (target = Warm) THENNo_Change Step 4: Obtain fuzzy valueFuzzy set operations perform evaluation of rules. Theoperations used for OR and AND are Max and Minrespectively. Combine all results of evaluation to form a finalresult. This result is a fuzzy value.Step 5: Perform defuzzificationDefuzzification is then performed according to membershipfunction for output variable.Consider the following real variables from everyday life:– Income measured in £UK.– Speed measured in meters per second.– A TV show measured in how much you are interestedwatching it.– A meal measured in how much you like to eat it.– A traffic light measured in what colour is on.In each case, suggest a fuzzy variable corresponding to these realvariables.For which of these five variables the use of a fuzzy variable isnot really necessary? Why?ActivityHints:I suggest the following fuzzy variables (you may come upwith a bit different):– Income: {Small, Medium, Large}– Speed: {Slow, Fast}– A TV show: {Boring, OK, Fascinating}– A meal: {Disgusting, So – -so, Good, Delicious}– A traffic light: {Red, Yellow, Green}It is not necessary to use the fuzzy representation for a trafficlight. The reason for that is that we only have to considerwhen it is either Red, Yellow or Green, and we do not need toconsider intermediate states.Video to watch:https://www.youtube.com/watch?v=rln_kZbYaWcThank You