FOLLOW US ON... © @brocodeengineeringofficial eS Brocode Engineering @brocodeengineering £) Brocode Engineering: For ® Engineering Studem> Educating People As Per New Scheme & Syllabus (AICTE FlexiblesCurricula) New Edition Vi Semester ——————— Machine LearningUNIT =I ion Inweesion Probability, statistics and ines! (earring, convex optimization: data visualization Fiypothesis function and Costing, data distributions, data preprocessing, data Augmentation, normalizing data sets Prechine learning models: Supervised and unsupervised learning to machine learning, scoPs and limitations, igebra for machine functions like sigmoid, il ‘on linearity, activation \ction, gradient descent, Linearity vs 1% 4 ReLU, etc., weights and bias, oss fun Re tayer network, backpropagation ; Weight initialization, training, testing, unstable gradient problem, auto encoders, batch ormalization, dropout, L1 and PF regularization, momentum, tuning hyper parameters. (94 to 108) UNIT - I ; ; Convolutional neural network, flattening, 2a ONL i olution layer, ooling layer, loss jayer, Fadaing, stride, convolution layer, pooling ye joni Gense layer 1x7 convolution ~ Inception network, input channels, transfer learnin learning, dimension reductions, implementation of tensor flow, keras etc. ig, one shot ‘ON like 20 to 124) UNIT -IV Recurrent neural network, long short-term memory, gated Peorrent unit, translation, beam search and width, Bleu scores attention model. Reinforcement learning, ‘equations, value iteration model, Q-learning, SARSA. 42510138) P, Bellman RL-framework, MDI actor-critic and policy iteration, aasto1®2) UNIT -V ‘Support vector machines, Bayesian learning, application of supine learning in computer vision, speech procese Te: natural language processing etc., Case study — ImageNet 160) competition. rasto" eee _ REERETT CTIGN YO MACHINE LEARNING, SCOPE AND Ti REGRESSICN, PROBABILITY, STATISTICS LINEAR ALGEBRA FOR MACHINE LEARNING, CONVEX OPTIMIZATION, ATA VISUALIZATION A Wa is machine tearning ors, Machine fearing is a branch oF seienee that deals with programmin the gvstems in such a way that they automatically learn and improve with experience. Here, learning means Feeo8! zing and understanding the input sererind making wise decisions based on the supplied data. this very difficult to cater fo all the decisions based on all possible inputs. ‘To tackle this problem, algorithms are developed. These algorithms build kowledge trom specie dole and past experience with the principles of statistics, probabil theory. logic. combinatorial optimization, search, -arning and control theory. ‘The developed algorithms from the basis of various applications such as~ & Vision processing (ii) Language processing ) Forecasting (e.g. stock market trends) (iv) Pattern recogn (v) Games vefPrrdER (vi) Data minin; e . (all Robutice, (vii) Expert systems ion BE. Writ 22 Mite the applications of machine learning. The applications of machi i T i chine learning are as follows — (Medical diagnosis fi ine Aki) Game oleying fii) Machine perception (iv) Information retrieval ivi) Affective computing -(viii) Recommender systems {X)_ Sequence mining 4v) Brain machine interfaces (ii) Natural al langua : Gx) Sem nnuage processingase. — eee (xii) Stock market analysis (xiii) Structural health monitoring (xiv) Syntactic. pattern recognition (xv) Computer vision, including object recognition, 0:3. What are machine learning tools ? Explain. we, Ans. Machine learning gives aset of tools that use computers to tana data into actionable information. Tools are a big part of machine learning choosing the right tool can be as important as working with the best algorthee Machine learning tools make applied machine learning faster, easier, Gos tools can automate each step in the applied machine learning process» shortening the time. J The machine learning tools are as follows — @ Platforms — Platforms are used to complete machine leaming project from beginning to end. (a) Provide capab learning project. (b) The interface may be graphical or command line (©) They provide a lose coupling of features. (4) They are provided for general purpose use and exploration rather than speed, scalability or accuracy. ies required at each step in a machine (ii) Library — Library gives capabilities for completing part of ¢ machine learning project. : (a) Provide a specific capability for one or more steps it machine learning project. ; (b) The interface is typically an application programms {etfate requiring programming (c) They are tailored for a spec! environment. (iii) Graphical User Interfaces — (a) Allows less-technical users use case, problem SP to work through machi " td learning. same . (6) Focus on process and how to get the most fr learning techniques. ; (c) Stronger focus on graphical presentati such as visualization. (@) Structured fo fons of inform e jeri process imposed on te user bY HE" (iv) Command Line Interface ~ (a)_ Allows technical users who are not programmers to work learning projects. (b) Frames machine learning tasks in terms of the and output to be generated. (c) Promotes reproducible results b: ‘commands and command line arguments. through mac] input required Y recording or scripting (») Application Programming Interfaces ~ (a) To incorporate machine learning into our own software projects. (b) To create our own machine leaning tools (©) Gives the flexibility to use our own processes and automations on machine learning projects. (a) Allows combining our own methods with those provided by the library as well as extending provided methods, (vi) Local Toots — Local tools can be downloaded, installed and run ‘on local environment. (a) Customized for in-memory data and algorithms. (b) Control over run configuration and parameterization (c) Integrate into our own systems to meet our needs, 1 (vii) Remote Tools — Remote tools can be hosted on a server and cal led from local environment. These tools are often referred to as machine learning as a service (MLaaS). (a) Tailored for scale to be run on larger datasets. b ‘ memory (b) Run across multiple systems, multiple cores and shared a wre short note on artificial intelligence vs machine learning. having nei Ncligence may be broadly defined as machines those intervention, Theo crY’ 2 Biven problem on their own without any human the necessary data sea cons HO! Programmed directly into the system but interpretation tay interpreting that data produce a solution by tse. Machine 80es underneath is nothing but a data mining algorithm. Providing the dats an, (8K°S Promote the approach to an advanced level by EXPOSEd 10 new doa ential FOF @ machine to train and modify suitably when information frome nits, 8 known as “training”. It focuses on extracting Underlying pattern nn eerOly large sets of data, and then detects and identifies ‘Meret new data wine Various statistical measures to improve its ability to Parameters Should bon resus more effective results. Evidently, some ‘uned” at the incipient level for better productivity. Aening iearning is the foot hold of artific togBign any machine having abilities as. nguage or vision, to get there at once. impossible to solve. Moreover, intelligent exposures. 1 Unit-l 7 cial intelligence tis ny ‘ , sociated with intelligence te gts Write the advantages of machine learning, That task would have been ain’ fas. The five advantages of machine learning areas follows ~ a system cannot be considered coma” ss ir is ir lacked the ability to learn and improne wen yea (Accurate ~ Machine learning uses data to discover the o its previg ‘on making engine for your problem. As you collect more data, the , ; , desacy can increase automaticaly, LDS Write and explain scope of machine learning (i) Automated As answers are validated or discarded, the machine ing mode can lear new patterns automaticaly. Thisallows users to embed a j , keine learning drecly into an automated workflow, ( Explaining Human Learning ~ mentioned eater, machin. ' : learning theories have been preceived fitting to comprehend features, of learn (ii) Fast - Machine nok can snare answers in a matter of in humans and animals. Reinforcement learning algorithms estimate jg milseconds as new data streams in, allowing systems to reac in realtime, dopaminergic neurones induced activites in animals during reward-tael (@») Customizable ~ Many data-driven problems can be addressed with learning with surprising accuracy. ML. algorithms for uncover machine learning. Machine learning models are custom built from your own sporadicdelineations of naturally appearing images predict visual fetumy data, and can be configured to optimize whatever metric drives your business, detected in animals initial visual cortex, Nevertheless, the important drivesa hursan or animal learning like stimulation, horror, urgency, hunger, inst actions and learning by trial and ergor over numerous time seles, ae note taken into account in ML algorithms. This a potenti opporniy di 9 wie me iadvantager of machine —— a more generalised concept of learning that entailsboth animals and macie Ans. The scope of machine learning are as follows we» ( Scalable ~ As your business grows, machine learning easily scales to handle increased data rates. Some machine learning algorithms can scale to handle large amounts of data on many machines in the cloud. Cc ining Machine Learniy Ans. The disadvantages of machine learning are as follows — iy ing ntaininy ine The disai ® : ~ eee are Tincee Mi algorithms are incorporated wit @. St learning has the major challenge called acquisition. Also Pret ES rere application software. The sede D4 on diferent algorithms data need to be proceed, Avy cee manually coded programs as pat of an application sofware TN SST puesed before providing as inpet io emnecane algorithms. Thus, it has a new programming language that is erat ee pesca Tange Significant impact on results to be achieved or obtained. “to be learned”. i i a subroutines as welll as those aie malig ‘use of this concept in: _) interpretation. That it result is also a Vike Python (Sckit-leam), R ete, ast Od as to develop amo! 106 5 he effectiveness of machine learning scope. But a fascinating new question is r id ines tagged as “to be leane. ii ; relevant learning experience for each subroutines B850 avingan), W© Say uses of machine algorithm is limited, Also, it’s not timing, and security in case of any unforesee re ae, tat it algoritns wil always work in every case imaginable, é , in most cases machine learning fails. Thus, it requires ; ter percero™® some under 1g it reqn function. nip Fordibn:A yentesacl concept eT ot eon une of the Problem at hand to apply the right algorithm. i Sa algorithms which are used in numerous Tt, Hi Taining dig gc ee? ating algorithm, machine learning also needs a lot ean Tink ieee in but nt mite 1 big One thoveh' PMs Amount of data Faye SBY if might be cumbersome to werk with a large peseentios = jg another potential er ee igi bea ‘ouch es Purposes, "tMMately, there are a lot of training data for image recognition recognition etc., is another Peer senses (€-8 Fol is the integration of different sens toestimate one problem is the in ich employ sel-supervised lar oe al psyche i! fea Bryjoltson ang prepare a system Wi hers, Researches in developmental Pasi fay nya ni knowledge using the others. ichine learning is its susceptibility to 7 at the; IcAfee said that the actual problem with this inevitable Jearning in humans when various a rests! AiTcute Ag peaney 0 make errors, ive leari ape Wee pate i Q noted more effective le “ihods in sinuate sim! S We have seen that i (°) One notable timitation of ma conn do diagnosing and correcting them can be ee : i “e Ht will need going through the underlying complexities. supplied, and studies on co-trsining ™ 98 Machine Leaming (Vi-Sem) (G8, Write short note on regression. WAS "Ans, Regression is used to map a data item to a real valued prediction variable, In actuality, regression involves the learning of the function that does, this mapping. Regression assumes that the target data fit into some known type of finction (eg. linear, logistic et.) and then determines the best function of this type that models the given data. Some type of error analysis is used to determine which funetion is “best”. Standard linear is a simple example of regression. For example, a college professor wishes to reach a certain level of ‘savings before his retirement, Periodically, he predicts whet his retirement ‘savings will be based on its current value and several past values. He uses ‘imple linear regression formula to predict this value by fitting past behaviour toa linear function and then using this function to predict the values at points jn the future. Based on these values, he then alters his investment portfolio, e229. Discuss straight-line and ‘multiple linear regression analysis. “Ans, Straight-line regression analysis involves a response variable, ¥ and a single predictor variable, x, Itis the simplest form of repression and models yas a linear function of x. That y= b+wx “where the variance ofys assumed to be constant, and b and w ae regression eine conte specifying the Y-intercept and slope of the line respective, The regression coefficients, w and b, can also be thought of as we! ights, so that we can equivalently write, yo wor WX “These coefficients can be solved for by the met estimates the best-fiting straight line as the one that minimizes th between the actual data and the estimate of the line. Let D be a raining: setconsising of ‘Values of predictor variable, x, for some population and their associated aes for response variable, y. The training set contains [D] dst poi oft fs 7 (x, ¥1)s a Yas» Opp Ypp- The reeression coefficients can be esti using this method withthe following equations — DB iu Loi-Oi-Y) shod of least squares, which | a Loi -¥ w= Fw : where & isthe mean value of x, X35» Xpand J sthemean valve ots vos Yip: The coeficints wg and wy often provide ‘good approximatio! otherwise complicated regression equations. Units 9 Multiple linear regression isan extension of straight-line toinolvemarethen one precio arabe Iallows raponsererale yest ‘modeled as a linear function of, say, n predictor variables or ites Ayan eur orcicm rete (310%, ¥2)-»» pp Yo), where the X, are the n-dimensional taining tuples with associated class labels, An example ofamulile linear regression model based on two predictor attributes or variables, Ay and Anis Y= Wo + WK) + WOR, 7 where xy and x; v i ere x, and xp are the values of attributes A, and Az, respectively, in X. 0.10. Explain regression and log-linear models. _Ans, Regression and log-linear models can b can be used to approxi given dat, in (simple) near regresion, the data ae modeled te ts ancighe line. For example, a random variable ae d le, y (called a response variable), modeled asa linear function of anoth oe er random variable, x (called a predi variable), with the equation ee vn y= wtb ere, the variance of y is assumed to numeri dattsestbuter, The cocicens wend ag oe peeseeremcetae coefficients, w and b are called regression minister af Sled fo by the method of eas squares, which of the line. Multiple line ‘actual line separating the data and the estimate FBeon, hich cha BESSON i an extension of simple) linear finction WS a response vari ‘ ctwoormee pean fe aie 10 be modeled as a linear -0g-linear m + distributions piaeat ‘approximate discrete multidimensional probability AMrbues) we cae OF UPS in n dimension (eg, described b Log inear meet onset each tuple asa point in an n-dimensional a ‘alidimensiont 5 "84 0 estimate the probability ofeach point in Subst of dimensgnarn er Seto! discretized attributes, based on a smal SPICE 10 become Combinations, This allows a higher sional dats : igher-dime A therefore ang agg Om lower mensional spaces. Loglnear models Regen for dimensionality reduction and data snothing ‘oplcnnes ny erutearmedels can both be used on sparse data although Meson diseacpin aa ite both methods can handle skewed data, appli ally wel i : i oie Vighdinarcnet Reeesson canbe compustinally tensive apt 10 soe TES logtnear meal show god sions, ir aeMachine Leaning (vi-Sem.) uth Mustated the simple linear regression and multiple linear sion derivation. fri in linear regression, the model specification is that the dependent «As Jr a tinear combination of the parameters need not be linear in Handenendent variables). For example, in simple linear regression for ve eling mijata points there is one independent variable— x;, and two parameters, Bo and By ~ Ye By Bist Ge ETE 2S me tn maple linear regesion, there are several independent variables or functions of independent variables. Adding aterm in x? to the preceding regression gives ~ yee Bo Buti t Baxi te t= 1,253) 0 Tis il lnearrearession; although the expression on the right hand side is quadratic in the independent variable x;, it is linear in the parameters Bo, By and B>- In both cases, observation. “0 18 an error term and the subscript j indexes @ particular Returning our attention to the from the population, we estimate sample linear regression model — i> B+ Bi The residual, e; =y;—Jj. is the difference between the value of the dependent variable predicted by the model, jj, and the true value of the dependent variable, y;, One method of estimation is ordinary least squares This method obtains parameter estimates that minimize the sum of squared residuals, SSE, also sometimes denoted RSS; straight line case ~ Given a random sample ‘the population parameters and obtain the SSE ye 1 an i ; Minimization ofthis funetion results ina set of normal equations 0 ; simultaneous liner equations in the parameters, which are solved 94 parameter estimators, fig,f, i ra Inthe case of simple regression, the formulas forthe least squares st ; Units Under the assumption that the population error term h 7 variance, the estimate of that variance is given by Poe a : n-2 This is called the mean square error (MSE denominator is the sample size reduced by the estimated from the same data, (n ~ p) for ) of the regression. The number of model parameters ») for intercept is used. In this ease p= 1 so the denominator isn DE ator isn = 2 The standard errors of the parameter estimates are given by E(x ~ T E(x, - 5)? 7 nce the further assumption that the population error term is normally ributed, the researcher can use these estimated standard errors to create confidence intervals and conduct hypothesis te e 01 uct is pothesis tests about the population Sy = <10 ee ee Fig. ‘ig. 11 Mustration of Linear Regression on a Data Set a ere is no ass Bredctor ofeach obrensutann ge, bonween and X(B = 0) he best sie prediction errors), is ¥ = fig (in terms of minimizing sum of squares —By ues - ney where isthe mean average) ofthe x values and isthe meat ©"! ZO ~¥), the on n this case, the total variation can be denoted as TSS => « fal sum of squares. ee — |Macnee ation between Y and X (B} #0), the best predictor ® 4 j,X; (in terms of minimizing sum of squares ‘error variation can be denoted as SSE = ‘When there isan associ ‘ofeach observation is Yj =Bo +BiXi ‘of prediction errors) In this ease, the i Sey, te ersor sam afsamares 7 sen TSS and SSE is the variation “explained” by the i The difference Dorwe having ignored X). It represents the regression of Y on X (as opposed to resent =LM-YY difference between the ited values andthe mean~ SSR x the regression sum of squares. ‘T9S= SSE + SSR - a lag La Fon WE Dey Ho? +L - VF i a Each sum of squares has a degrees of freedom associated with it, The total degrees of freedom is dfyaqy = 0 - 1. The error degrees of freedom is feqoe= 2 ~2 (for simple regression). The regression degrees of freedom is (for simple regression). fjouat = Meroe + Hregression n-I=n-2+1 , ‘Table 1.1 Analysis of Variance for Simple Linear Regression [Source Taf] SS | MS Fob [ le Rearesion | | sop =3 = SSR | py, = MSR | pci g 22 Fo) (odely an Zs | MSR=A | Fabs =fagp | PFin-2 2 Fas wse= S56 Trained “ a Tota! a) nA) Ts8-S(y,— (Coreey || 2M-¥ Error and repression sums of squares have a mean square, which is ¥f sum of squates divided by its corresponding degrees of freedom~ M ~ SSE(n ~ 2) and MSR = SSR/1. It can be shown that these mean sat" a 7 have the following expected values, average values in repeated sampliné the same observed X levels — E(MSE} = 0%, E(MSR} = 7 +B7 (Xi XV” a : sion Q.13. Explain analysis of variance for multiple linear regres Xp (Pr yer’ * fy cin ‘Ans. When there is no association between Y and X 0), the best predictor of each observation is Y minimizing sum of squares of prediction errors). In this case, the total variation can be denoted as TSS = J°(¥; —¥)°, the total sum of squares, just as with simple regression. ba When there san association between ¥ and at leastone of X, Xp (not all B,=0),the best predictor of each observation is 0 FBX +... + ox, (inten of minimizing sum of squares of prediction eons). Inthiscae ten, variation can be denoted as SSE= 3°(Y; ~¥5)°, the ror sum of squares ‘The difference between TSS and SSE is the variatic regression of Y on X;, ion “explained” by the vv» Xp (@S opposed to having ignored X;, .. Xp). It represents the difference between the fitted values and the mean— SSR = DC% -¥)? the regression sum of squares. a TSS = SSE + SSR » " 2 DG -¥Y = Dy - hy? + D0 -¥)? a a a Each sum of squares has a degrees of freedom associated with it. The total degrees of freedom is dfy,.,) = n ~ 1. The error degrees of freedom is errr = Ap. The regression degrees of freedom is dfycsresion =P. Note that when we have p= I predictor, this generalizes to simple regression. rota = Werror + Al Represson n-1=n-p'+p Error and regression sums of i - squares have a mean square, which is the ce Squares divided by its corresponding degrees of freedom- MSE. have the folowing MSR ane It can be shown that these mean squares expected v: lues it the came sonore pects t ‘alues, average values in repeated sampling at E(MSE)= o?, E{MSR} > 0? te 1.2 An: Ysis of Variance for Multiple Linear Regression a a WF | se 5 [ry MSE haMachine Learning (¥-2"'" t3$)4. What is the concept of pro 4, What em “The concept of probability is extremely important, It has found q Ans. ' bi The rery extensive application in the developmen! of all physical chance ‘of happen ing of an event when expressed quantitatively is call 1 probability. ‘Theorem ability ? Explain with theorem, of Total Probability or Addition Law of Probability — 1F the probability of an event happening as a result of a trial is PCA) and the probability of a mutually exclusive event B happening is P(B), then the probability of ether of the events happening as a result of the trial is P(A +B) or P(A w B) = P(A) + PCB) Proof. Let n be the total number of equally likely cases and let m, be favourable to the event Aand my be favourable to the event B. Then the number of cases favourable to A or Bis my + mp. Hence the probability of A or B happening as a result of the trial__ = MAM TL ays pee n a oA If the events A and B are not mutually exclusive, then, there are some outcomes which favour both A and B. If ms be their number, then these are included in both m; and my, Hence the total number of outcomes favouring either A or B or both is my +m; —my ‘Thus the probability P(A +B) or P(A UB) of occurrence of A and B or Both m3 n or P(A + B)= P(A) + P(B) ~ P(AB) or P(A UB) = P(A) + P(B) — P(A 4B) ‘when A and B are mutually exclusive P(A 7B) or P(AB) = 0 and we have P(A +B) or PCA UB) = P(A) + PCB) Proved Particular Cases — (© IfAand B are defined on the sample space S, then (A UB)= n(A) + (8) __ PAUB)= P(A) + P(B) () Since S and @ are mutually exclusive events and SU ¢=5 PSU 4)= P(S) PS) + PC) = P(S) (iii) Let A and A be complementary events, Then A and A are mutually exclusive by definition. Hence P(A A)= P(S) P(A) + P(A, PCA) = 1~P(A) or P(A)= 1-P(A) (iv) We know that [ry P(S)=1) A= (ANB)U(AnB) That A is the union of two mutually exclusive events P(A)= P(ANB)+ P(A B) or P(AMB) = P(A)-P(AMB) or P(AMB) = P(A)- P(AMB) Similarly, P(ANB) = P(B)—P(ANB) (¥) IfBCA, then (@) P(AMB) = P(A)—P(B) (b) P(B) < P(A). QS. Define the following ~ @ Probability function (ii) Probability mass function. (RGR, June 2016) ass dns. Probability Function — In terms of symbols of a variable X can lime discrete set of values X,, Xp, ... Xj With respective probal ae where Pi* Po +..... + p= 1, we say that defined er obability distribution for X has been [A ned. The function P(X) which has the x mettive values py, po... for X = Xp, X, Xs called t it function oe LeProbabitity function or frequency Fig. 1.2 In probability theory and statistics, ction that gives the probability that (D Probability Mass Function—Machine Leaning (ViSem.) mass function differs from a probability density function fin tatthe laters associated with continuo rater than discrete random (ra the values of the ater are not probabilities s such: pd. musty integrated over an interval 10 yield a probability. Q.16. Write the expected value (or mean) and variance of a discrete example. probability distribution with : dns. Since the probability distribution for arandom variable isa model for bution of a population (think of organizing a very the relative frequency distri (thi ‘of discrete random variable into arelative frequency large number of observations a dieribution the probability distribution would closely approximate this), the analogous descriptive measures ofthe mean and variance are important concepts “The expected value (or mean), denoted by E(X) or p, of a discrete random variable X with probability function p(x) is defined as follows — QQ = Dx Pe) where the summation extends over all possible values x A probability 1 Note the similarity to the mean =a fx(wheren = Ef) of a frequency distribution in which cach value, x, is weighted by its frequency f. Fora probability distribution, the probability p(x) replaces the observed frequency f. The variance of a random variable X, denoted by Var(X) or 0”, is defined as the expected value of the quantity (X — .)? where jis the mean of X, i.e. VartX) = E{(X =p)*] = D0x-w)? pO) ‘The standard deviation of X is the square root, o, of the variance. of tossing a fait For Example — Consider an experiment that consists 7 el coin three times, and let X equal the number of heads observed, deseribs probability distribution of X and hence determine E(X) and Var(X) There are 8 possible outcomes here - HHH, HHT, HTH, THH, H THT, TTH, TTT. This gives probability distribution as follows - (rT. X| Outcomes | Probability 0 TIT ve 1) RTT,THT,TTH | 3/8 2.) HHT,HTH,THH | 3/8 3 HHH V/s Total Probability | 1.0 _| 2 Fane in determi ow there are ma pad Some of these Ismomial funeto Units 47 ‘The expected value of X is— E(X) = O(1/8) + 1(3/8) + 2(3/8) + 3(1/8) + 3/8 + 6/8 + 3/8 1218 =15 The variance of X is Var(X) = (0 - 1.5)°(1/8) + (1 ~ 15/8) + 2 - 1.5°G/8) a + B= 15) 2.25(1/8) + 0.25(3/8) + 0.25(3/8) + 2.25(1/8) ee (2.25 + 0.75 + 0.75 + 2.258 =68 =0.75 There is an alternative way of computin; ting Var(X). It can be shown that — Var(X) = Ex? p(x) - [Ex p(x)}?. Using this result we have the table 1.3, Table 1.3 =| pls) | xp(x) | x? pee) 0 8 0 0 1 3/8 | 3/8 ws 2 3/8 | 6/8 | 12/8 3 V8 | 3/8 9/8 Total | 1.0 | 15 3.0 ‘Var(X) = 3.0 - 1.5? = 3.02.25 15. 047. What do you mean iepldsaaiaa, 7 probit eae you mean by probability dsribuion and discrete edcssedae Distribution — We have explored the idea of probability, the variable betog cs ongePt oF probability distibution. In situations where by a probatilgae Suied is arandom variable, then his ean often be modelled eaaponeaee! 4 distribution, ‘Simply put, a_probat is tic com ihe collection of possible values that the variable can take, Probability that each of these values (or a subset of these we together with 1 Values) oceurs, So, in stochas inten Probability distribution is the equivalent of dere line Gin which theres no uncertain). As we Pia pans ons available for deterministic modelling 8 and exponent ch an others eg. linear functions. ‘unctions. Exactly the same situation18 Machine Leaning (VI-Sern,) prevails in stochastic modelling. Certain probability distributions are used more Bhen than others. These include the Binomial distribution, the Poisson distribution, the uniform distribution, the normal distribution and the negative exponential distribution. Discrete Probability Distributions — A discrete random variable assumes, each of its values with a certain probability, each possible value of the! random variable has an associated probability. Let X be a discrete random Variable and leteach value of the random variable have an associated probability, denoted p(x) = P(X = x), such that D1 [en P(X): | Pi | P2 | + f+ | + | Pn ‘The function p(x) is known as the probability distribution of the random variable X if the following conditions are satisfied — (pls) 2 0 forall values x of X Gi) Loo p(x) is also referred to as the probability function or probabi (948. What do you mean by probability density function ? IR.GRV., May 2019 (I-Sem.)] Ans. A discrete random variable, probabilities are associated with particular individual values of the random variable and the sum of all probabilities is ‘one. For a continuous random variable X, we do not have a formula which gives the probability of any particular value of X. The probability that 4 continuous random variable X assumes a specific value x is taken to be zero. For a continuous random variable we deal with probabilities of intervals rather than probabilities of particular individual values. The probability distribution of continuous random variable X is characterized by a function f(x) known as the probability density function. This function is not the same as the probability function in the discrete case. Since the probability that X is equal a specific value is zero, the probability density function does not represent the probability that X =x. Rather, it provides the means by which the probability of an interval can be determined. ‘The function f(x) whose graph is the limiting curve is the probability density function ofthe continuous random variable X, provided that the vert? scale is chosen in such a way that the total area under the curve is one Let X be a continuous random variable. The function £{x) is said 1°" Probability density function, if mass function. i Unit.) 19 (i) f(x) 20. for every value of x Gy FF food ity P@SXSd)= fFCOUs and for any a and b Note ~ The third condition indicates how the function is used ~ ‘The probability that X will assume some value in the interval [a, b] =the area under the curve between x =a and x = b. This is the shaded area as shown in fig. 1.3. Probability Pasxst) Density Function In general the exact evaluation of areas requires us to use the probability density function f(x) and integral calculus. These calculations are time consuming and not straightforward. However, probability values can be obtained from statistical tables (just as for discrete probability distributions). 0.19. Write short note on statistics. Me so ark Te calculation involving the ratio of sizeof difference (numerator) {Gameunt of error (denominator, illustrated in the table, produces a number eraraauisti. ts F. and 2? are statistics. This value becomes larger a8 the A larger gnegen&® (ot relationship to error increases asthe formulae indicate ic obviously is better. enought it compared to a table of statistic to determine if it is large need not be ie null hypothesis. As the sample size increases, the statistic larger samples, irge to reject the null hypothesis. The reason for this is that Unrepresentatie oe nore confidence that the obtained sample is not Sotaiouly teeth, Papulaion that is, nota luke". Smal samples are Population will hae eeause every small sample that is drawn from a {bat the ratio ig ene different due to chance. Therefore, to be confident ‘Small, the statist ‘ciently large to reject the null hypothesis when the sample naan Test Be ager Weneed fie Would leam the several formulae used to calculate ler to determine if we can reject the null hypothesis.rd 20 Machine Learning (VI-Sem.) A few of these formulae are presented here to point out their similar; Conceptual, not actual, formulae are used. The different types of formula shown in table 1.4. ities, C ate Table 1.4 Name of Test Formula (Conceptual) What is Does | fa Mean difference Difference ee Error variance two means | Analysis of F Differences among means aaa among | ii eee ans, wit | variance Error variance within groups many IVs Extent to which frequencies are not _| Differences in Chi-square consistent with the null hypothesis _ | frequencies; or test Size of sample relationship between| nominal variables Q.20. Explain the different types of statistics and compare them with example. re Ans. There are two major types of statistics — (i) Descriptive statistics (ii) Inferential statistics. (i) Descriptive Statistics — The branch of statistics devoted to the summarization and description of data is called descriptive statistics. Descriptive statistics consists of method for organizing and summarizing information. Descriptive statistics includes the construction of graphs, charts, and tables, and the calculation of various descriptive measures such as averages, measures of variation. and percentiles. In fact, the most part of this course deals with descriptive statistics. (ii) Inferential Statistics — The branch of statistics concerned with using sample data to make an inference about a population of data is called inferential statistics. Inferential statistics consists of method for drawing and measuring the reliability of conclusions about population based on informatio" obtained from a sample of the population. : Inferential statistics includes methods like point estimation, inters? estimation and hypothesis testing which are all based on probability theo" ider even! e he samp! Jowil Example (Descriptive and Inferential Statistics) — Cot 5 tossing dice. The di rolled 100 times and the results are forming th ' data. Descriptive statistics is used to grouping the sample data to the fo table 1.5.22 Machine Learning (Vi-Sem.) of Inferential staisties provide @ wa going from a “sample” to a™ ‘population” inferring the “parameters” of population from data on the of a sample, i.e. parameters like j-and o. from statistics like m and s But before we ean see what is involved in the move from sample yy population we need to understand how to move from population to sample, The study of obtaining a sample from a population is “probability”, —= Probability — Population [Sample =— Inferential statisties <= ty of picking a black ball from jar A is one For E: half; the probability of picking a black ball from jar B is one tenth, ample ~ The proba jar A~ | Jab B | 50 black, | 90 black. | 50 white | 10 white 4.22. Explain the term statistical analysis of data.(R.GP:, May 2019) Ans, Statistical data analysis isa procedure of performing various statistical operations. It is a kind of quantitative research, which seeks to quantify the data, and typically, applies some form of statistical analysis. Quantitative data basically involves descriptive data, such as survey data and observational data. Statistical data analysis generally involves some form of statistical tools, which a layman cannot perform without having any statistical knowledge. There are various software packages to perform statistical data analysis. This software includes Statistical Analysis System (SAS), Statistical Package for the Soci! Sciences (SPSS), Stat sofi, etc. Data in statistical data analysis consists variable(s). Sometimes the data is univariate or multivariate. Depending up" the number of variables, the researcher performs different statistical technique If the data in statistical data analysis is multiple in numbers, then seve! ‘multivariates can be performed. These are factor statistical data analy discriminant statistical data analysis, ete, Similarly, if the data is singula number, then the univariate statistical data analysis is performed. This incl” ‘test for significance, z-test, f-test, ANOVA one way, etc. “The discreet data is the one that can be counted. For example, the ‘bulbs can be counted. jous data in statistical data analysis is distributed under can also be called the probability density counted number 0} The continu continuous distribution function, function, or simply pdf. “The discreet data in statistical data analysis is distributed under discreet distribution function, which can also be called the probability mass function or simple pmf. We use the word ‘density’ in continuous data of statistical data analysis because density cannot be counted, but can be measured. We use the word ‘mass’ in liscrect data of statistical data analysis because mass cannot be counted. ‘There are various pdf's and pmf’s in statistical data analysis. For example, Poisson distribution is the commonly known pmf, and normal distribution is the commonly known pdf. These distributions in statistical data analysis help us to understand which data falls under which distribution. If the data is about the intensity ofa bulb, then the data would be falling in Poisson distribution. There is a major task in statistical data analysis, which comprises of statistical inference. The statistical inference is mainly comprised of two parts — estimation and tests of hypothesis. Estimation in statistical data analysis mainly involves parametric data — the data that consists of parameters. On the other hand, tests of hypothesis in statistical data analysis mainly involve non Parameters data ~ the data that consists of no parameters. Traditional methods for statistical analysis ~ from sampling data to erreing results ~ have been used by scientists for thousands of years. But am i volumes make statistics ever more valuable and powerful. ominene ne, Powerful computers and advanced algorithms have all led of computational statistics. Wheth Permtations ornare WoFking with large data volumes or running multiple for today's sao 0 is a fixed parame wecueeration 4s that (x) represents the cost of using x on the it* elem" “Simplicity gee ate ROO is a regularization term which enforces som 'y" in x. We discuss now ‘major instances of equation (i). In all ‘one has a data set of n cos! linciion ¢ dosent, the form (w;, yi) © R*«y,i=1,2,3, ...., mand Units) 34 In classification one has Y = {-1, 1}. Taking (x) = max(0, I — y,xTw,) (the so-called hinge loss) and R(x) = ||x|If one obtains the SVM problem. On the other hand taking f(x) = log(1 + exp(-y,xTw,) (the logistic loss) and again R(x) = ||x|[3 one obtains the logistic regression problem. In regression one has Y = R. Taking f(x) = (x"w, —y,? and R(x) = 0 one obtains the vanilla least-squares problem which can be rewritten in vector notation as — min || Wx- ||} xcR" where W © R'™* is the matrix with w7 on the i row and Y= (yy... Yq) With R(x) = ||x|[3 one obtains the ridge regression problem, while with R(x) [xl, this is the LASSO problem. In our last example the design variable x is best viewed as a matrix, and thus we denote itby a capital letter X. Here our data set consists of observations of some of the entries of an unknown matrix Y, and we want to “complete” the unobserved entries of ¥-in such a way that the resulting matrix is “simple” (in the sense that it has low rank). After some. ‘massaging the matrix completion problem can be formulated as follows — min, Tr(X) st Xe R™, XT=X, X30, x, where Q < [nP and (¥;.)¢ eq are given. yr Y,, for(i,j) (40.31. What do you understand by data visualization ? Discuss some Python's data visualization tools suck as box plots pie charts and bar charts in brief, Ans. Data visualizations helps top management who are the decision makers to view analyties being visually represented, so it makes them tocasily understand the complex ideas and identify the new structures or patterns, When t'sualization becomes interactive, then we are able to push the concept alittle further thereby using technological tools to grasp more details from graphs and charts, therefore making changes to the data that is being seen and how Such data is being processed. metnoa Treats Putting data forward and representing them in a particular About in eect Which contains some variables and attributes for bringing incarnation. Visualization-based data discovery techniques gives room make up sources of completely differen jn32. Machine Leaming (Vi-Sem.) (0 Box Plots— A box plot i a graphic representation of quanti variables based on their quartiles, as well as their smallest and largest y The simplest box plot contains a box and two whiskers. The bottom and tonay yg ales, the box are the first and third quartiles, and the Tine inside the box is the mediay The box area represents the range between the first and third quartiles. whieh is called the interquartile range (IQR). The ends of the {wo whiskers indicay the maximum and minimum values of the variable. Box plots can have ma variations. For example, complex box plots mark outliers (three or more time, the IQR above the third quartile or below the first quartile) and suspect outliers (one and a half or more times the IQR above the third quartile g below the first quartile). Ina complex box plot, if either type of outlier appeas, the end of the whisker on the appropriate side changes to one and a half 1QR from the corresponding quartile. The end of this whisker is defined as an inte: fence, and the third IQR from this quartile is considered as the ovter fence. ‘Outliers in this plot are displayed as filled circles and suspected outliers are displayed as unfilled circles. Figs. 1.4 and 1.5 show an example of simple ad complex box plots. A Manion 1s wt | 4 outers wp | 4 " 1 outer Fea ‘Third Quartile Suspected [5 wel 4 mies Fg fusion Inner Feat Median ae 5 | 1oR int Quartile ano ioe First Quart of Le minimus ° 4 Fig. 1.4 Simple Box Plot Fig, 1.5 Complex Box Plot We can see that it’s super easy to create this plot with Matplotlib. All "€ need is the function plt, boxplot( ). The first argument is the data points The simple box plot in using python's values ~ [1, 2, 5, 6,6, 7,7, & 8,8, 9, 10, 20} pliboxplot values) plt.yticks(range(, 21)) pit ylabel(*value") pltshow( | (ii) Pie Charts ~ tis as well-known ph. A pic chartshows information a way that is not difficult fo toad called “pic-slice” Form and the various sues ofslice show's how much of an element is inexistence. When the slice is big. then it hows of the data was gathered. It is also used to compare values of data and the moment some values are represented on pie chart, then we will be able to view which of the items is the least popular or which is more popular. The best and effective way to make use ofa pie chart is when statistics and data i report consumers do not have to gues Unit-) 33 Fig 1.6 A Simple Standard Pie Chart they contain a few components and when the percentages and texts are also involved in order to detine the content. By providing additional information, the meaning and value of each slice. If you choose to use a pie chart, the slices should be a percentage of the whole. Donut Chart Demonstration tater Tal = Lbicemes 12597 whites raat diel nae % S.Insurance 5.59% = 6 Facittes roses + 7 Production 2.8L vp B Ble ery hey wm Puunphia 1 Peca Hig. Li A Simple Exploding Pie ChuctA wedge is used to representa data parts that has the sa and the pie chart control usually decides the data wedge sive rari compared with the other data wedges. Pie charts consist Dough are almost same as the standard pie chart just that it consist of hollow a and the exploding charts, the wedges ae being obtained from the othe wedge Apie chart with Matplotlib, we can use the plt.pie() function, The ayy parameter allows us to display the percentage value using the Python se: formatting. ‘ sizes = [25, 20, 45, 10] labels = ["Cats', "Dogs’, "Tigers", "Goats"] plipie(sizes, labels=labels, autopet=""%4.2!") pli.axes( ).set_aspect("equal") pltshow( ) (iii) Bar Chart — Bar chart is as well referred to as column chart ai they are used to for comparison of items of different groups. The bars are us! to represent the various values of a group and the bar chart makes use of bet horizontal bars and vertical bars. When the values to be represented are clea different and such differences in the bar are been seen by human eye, then os can decide to make use of a bar chart, but when there are very huge numbss of value to be displayed, then it might be a bit more hard to make compari between the bars, Most times, bar chart is used to represent discrete data oat is as well used to present single data series while the data points that are are often being grouped in a series. ‘The Cookie Shop 2003-2005 Income 90000 80000 70000 60000 50000 40000 30000 20000 10000 so 200320082005 Fig. 1.9 Displaying a Simple Bar Chart tion ‘A bar chart with Maplotlib, we will need the pi-bar( ) Funct! ‘#Our data, when its ben om “aan of two popu variations called Doughnut chart and Exploding pie chart. The . Unit-l 35 tabels = ["2003", "2004", "2005"] usage ~(0, 10000, 20000, 30000, 40000, 50000, 60000, 70000, 80000, 90000) #Generating the y positions. Later, labels. y_positions = range(len(labels)) #Creating our bar plot pltbar(y_positions, usage) plt.xticks(y_positions, labels) pltylabel("Usage (%)") pi.ttle("The cookie shop") pH show( ) we will use them to replace them with \Q32. Explain data visualization techniques. ‘Ans. Visualization is the use of computer-supported, visual representation of data, Unlike static data visualization, interactive data visualization allows, users to specify the format used in displaying data. Common visualization techniques are as shown in fig. 1.10. Line Graph Pie Bar char Data Visuatication | Bar, Scatter Plot Fig. 1.10 Commonly used Data Visualization Techniques @ Line Graph ~ This shows the relati it ionship between ite it to compare changes over a period of time. ms Hea " ‘Bar Chart~Thisis used to compare quantities of dife:ent categories. (ii) Scatter '— This is i ‘ of two ite Seater Plot — This isa two-dimensional plot showing variation (iv) Pie Chart — Thus, the format of chart, fin line graph, ete, Its i for your ehh et. Is importa be used This is used to compare the parts of a whole. ‘graphs and charts can take the form of bar chart, pie int 0 understand which chart or graph to use~y Data visualization uses comp Data nuter graphics to show relationship among elements of the data. can generate scatter plots, and other types of data graph with sim mouse clicks. Colors are carefully selected for ¢¢ When color is used to represent data, we differentiate between data elements. In data visualization, data is abstracted and Summarized. Spatial vari, such as position, size, and shape represent key elements inthe we” Visualization system should perform a data reduction, transform and st the original dataset on a screen. Pes It should visualize results in the form of charts and results in user friendly way. Patterns, trendg Pie charts, bar gh le pull-down menue ertain types of visu; alia must choose effective egiae™ Clos ‘graphs and prese, “£933. Discuss the application of data visualization. Ans. Most visualization designs are to aid decision making and serves tools that augment cognition. In designing and building a data visualization prototype, one must be guided by how the visualization will be applied. Das visualization is more than just representing numbers; it involves selecting ant rethinking the numbers on which the visualization is based. Visualization of data is an important branch of computer science and his wide range of application areas, Several application-specific tools have beet developed to analyze individual datasets in many fields of medicine and science @ Public Health — The ability to analyze and present data in at understandable manner is critical to the success of public health surveillance Health researchers need useful and intelligent tools to aid their work. Secun? is important in cloud-based medical data visualizations. Open any medial health magazine today, and we will see all kinds of graphical represental Si compared (ii) Renewal Energy ~ Calculation of energy consumption compa to production is important for optimum solution. fe Gi) Environmental Science — As environmental manta required to make decisions based on highly complex cate they hae visualization. Visualization applications within applied environ are beginning to emerge. It is desirable to have at one’s dispos programs for displaying results. vee alization ™ ule" ortant (iv) Fraud Detection — Data visualization is importan Fs estigator may use data vist vestigation. Fraud investigator may val stages of fraud investigation. Frit iv ieewte patterns that suBeest Unit-1 37 (») Library-decision Making ~ Data visualization software allows librarians the flexibility to better manage and present information collected from different sources. It gives them the skill to present information ina creative, compelling way. Visualization of library data highlights purchasing decisions, future library needs and goals. Librarians, as de facto experts of data visualization, can assist students, faculty and researchers visualize their data, Several information visualization algorithms and associated software have been developed. These software enable users to interpret data more rapidly than ever before. Write short notes on the following — nar Histogram (ii) Quantile plots (ii) ¢-4 plots (iv) Scatter plot (v) Loess curve. Ans, (i) Histogram — Plotting histograms, or frequency histograms, isa graphical method for summarizing the distribution of a given attribute. A histogram for an attribute A partitions the data distribution of A into disjoint ‘subsets, or buckets. Typically, the width of each bucket is uniform. Each bucket is represented by a rectangle whose height.is equal to the count or relative Frequency of the values at the bucket. IFA is categorie, such as automobile ‘model or item_type, then one rectangle is drawn for each known value of A, ‘and the resulting graph is more commonly referred to as a bar chart. If is ‘numeric, the term histogram is preferred. In an equal-width histogram each bucket represents an equal-width range of numerical attribute A. Fig. 1.11 shows a histogram for the data set of table 1.6, where buckets are defined by equal-width ranges representing $20 increments and the frequency is the count of items sold 000 im Ee 000 3 ro < a 40-597 00-79 w0-09 Toh 1ST LT Unie Prices)38 Machine Learning (Vi-Sem.) ‘Table 1.6 A Set of Unit Price Data for Items Sold ata rang All Electronics Anch ot Unit Price (8) Count of Items Sold 40 275 43 300 47 250 ( Gi) Quantite Plots ~ A quantile plot is a simple and effective ways have a first look at a univariate data distribution. First, it displays all ofthe data for the given attribute. Second, it plots quantile information. The ‘mechanism used in this step is slightly different from the percentile computation Let x;, for i= 1 to N, be the data sorted in increasing order so that x, is the smallest observation and xy is the largest. Each observation, x, is paired wit a percentage, f;, which indicates that approximately 100 f; % of the data ar below or equal to the value, x;. We say ‘approximately” because there may not bea value with exactly a fraction, f, of the data below or equal to x;.Fig, 1.12 shows a quantile plot for the unit price data of table 1.6. 140 . 120 sim on 80 wee” oo . En feseoeet 20 ° 0.000 0280 070 1.000 vale Fig. 1.12 A Quantile Plot for the Unit Price Data of Table 1.6 the Gil) q-q Plot — A quantite-quantile plot or 4-4 plol, caiee quantiles of one univariate distribution against the corresponding at" another. It is powerful visualization too! in that itallows the user| there is a shift in going from one distribution to another. S Fig. 113 shows quantile-quale plot forunit price dats oF ems ‘ { branches during a given time period. Exch point co ns sold? suwo differ \ . ther attribute Unite 39 et and shows the unit price of items sold at wuantile for each data set and dit price at each branch is the same, . In addition, the darker points correspond to the data for Q,, the median, and Qa, respectively. Branch 2 (Unit Prices) - “oso 60 70 80 9) 00 HO a0 Branch 1 (Unit Price) Fig, 1.13 A Quantile-quantile Plot for Unit Price Data from Two Different Branches (jv) Scanter Plot - A scatter plot is one of the most effective graphical methods for determining if there appears to be a relationship, pattern, or trend between two numerical attributes. To construct a scatter plot, each pair of values is treated as a pair of coordinates in an algebraic sense and plotted as Points in the plane. Fig. 1.14 shows a scatter plot for the set of data in table |,6. The scatter plot is a useful method for providing a first look at bivariate data to see clusters of poi i see clus Points and outliers, or to explore the possibilit correlation relationships. mone te possinny of 3 ose See z ~, - z oe > 7 40 Go 80000140 Unit Price (8) when of tae A Scatter Plot for the Data Set of Table 1.6 Exension io the eaten yee weet attributes, a scattersplot matrix is a useful OF scatter plots thay ony oeR N attributes, a scatter-plot matrix ig ane | ‘st Provides a visualization of each attribute with every _ i ae40 Machine Learning (VI-Sem.) () Loess Curve— A loess curve is an imy : portant explorao aid that adds a smooth curve toa scatter pltin order to provide bere ee of the pattern of dependence. The word loess is short for “local rane Fig. 1.15 shows a loess curve for the set of data in table 1.6. SS". ems Sold 0 2 40 60 80100120140 Unit Price (5) 1.15 A Loess Curve ‘Two parameters are needed to fita loess curve. a, a smoothing paramete, and A, the degree of the polynomials. While & can be any positive number ‘2.can be | or 2. The goal in choosing «is to produce a fit that is as smoothas possible without unduly distorting the underlying pattern in the data. The cunt becomes smoother as a increases. There may be some lack of fit, howeve. indicating possible “missing” data patterns. If a is very small, the underly pattern is tracked, yet overfitting of the data may occur where local “wiggle in the curve may not be supported by the data. If the underlying patter oft data has a ‘gentle’ curvature with no local maxima and minima, then local linear fitting is usually sufficient (2.= 1). However if there are local maxi minima, then local quadratic fitting (A = 2) typically does a better job! following the pattern of the data and maintaining local smoothness 0.35. Write short notes on the following — (@ Table chart (i) Bubble chart (iii) Tree map (iv) Parallel coordinate (v) Line chart (vi) Area chart. ‘Ans. (i) Table Chart ~ Table is simply the arrangeme! and column. In conducting research and analysis of data, ihe very important. Tables are simple to understand and analy7® 0° interpret the method of data representation. A row is represelie and column is as well 2 representation of records that have er times, this order of arrangement can be changed, ©. (OMS Tr gigas aanris and columns represent variables. Average annual 7546 qj 2 elected categories of consumer spending, by housing lene are shown in table 1.7. ant of data using" role cof table " it's sino jon of vaio 4 of values Pe Unit) 41 Renter Percent on) Percent rome | 2010 [Peent| 1986 | 2010 | Change Tam | ETH] 0 SSA] Srr| anno, 1a |” steo|aa02) 2 Asan] ooo] 13 | 2972 : 3327| 220 3 reas? | 18503 16 1341 ” Aeon 9036 31 ang ae toe 2458 18 Gent ard 2a " 234 124 inment 3,088 ~~ Peta inranc and pensions bs 2 (i) Bubble Chart — A bubble plot is some degree of difference of a seater plot and the markers init are being substituted with bubbles and this is possible only we have a set of data points which has three values contained in cach data item. It shows the relationship that exists between the minimum of three variables. Two of them gets represented by the plot axes i.e., x-axis and y-axis, while the third one by the bubble size and each bubble isa representation of an observation Bubble plot is used with a lot of value, say hundreds of them or also used ithe values are somewhat different by numerous structure of magnitude. Colors ing pring used to represent an additional measure and the bubbles could be "ected to animation in order to show data changes over a period of time. so Annual Sales Chart : ~ oO 7 1 %eoe 9 4s Origin Cost (Bg) 25 30 35 al bbb pg Fe 1.16 Simple Bubble Plot 'Neoftisk and success neat ete in project mana esi ‘volved in executing.a proje igement in comparing set and where there are42, Machine Leaming (Vi-Sem) values, then the probability of success and the tj three values as net pres sum represent the bubble size. (ii Tree Map ~ Aree map is a visualizing technique that has he attribute of showing data in hierarchy in a nested or layered rectangle form, jaa very effective technique that is used to visualize structures of hierarchic, User are able to compare nodes and sub nodes at different depth and also they are able to identify expected results and patterns, A lot of data set have the hierarchy characteristics and the objects are thereby divided into differen divisions, sub divisions, ete. Immune System ‘Non Specific Defenses Specific Defenses quired Prevents any Pathogen Immunity Hummoral Cell Meditated Immunity Lmmunity First Line Defense Second Line | | Defense Bells Tce: ‘Skin Prevents Pathogen from Entering Body Interferons exer Inflammatory Response ‘Mucus, Tears, Saliva Redness and Swelling Hairand Cilia Increase in Blood Flow WBS (White Blood Cells) Eat Up Patoges (Pagsest©) Stomach Acid Histamines Signal Fig. 1.17 Tree Map Display Hierarchical Data () Parallel Coordinates — The parallel coordinate technique mek use of the concept of networking a multi-dimensional point to some axes * all ofthese are parallel to each other. In these technique, single data eleme are being plotted across many dimensions and these dimensions are cones toa y-axis and each object of the data is shown along the axes as @ series Sennected pont. The parallel coordinate is important if you want io so on roneansional data and a lot of these dimensions are being organ” Net this technique. fe sean is a line that forms a single polygonal line for ie ‘Therefore qa esemted, then it connects the individual coordinate map pie ALThinvisueet OF dimensions that is being represented i8 NOt HIV Geom ngue is applicable in areas such as computer ¥* ‘ional geometry, robotics and data minin&: Units 43 sdadvantage of this visualization technique is that it usually represents Ago sons of dimensions without limits. Though, you can encounter a ease such as von zonal Fines being overlapped which causes difficulty in identifying the data and this caused when you have many points that are characteristics in ch. feig represented when engaging the parallel coordinate appr (») Line Chart — Line chart is common well known graph in many fields, also label as line graph. It isa graph which is use to display information ‘connected points, These points are connected through continuous or straight line. Line graph is the extension of scatter plot. Data points can be represented by icons or symbols, or can also draw simple line without icons. The line chart is often used to visualize a trend in data over time interval, means it use to show tendency in the data set and illustrate the data behaviour withthe passage of time or over a specific interval of time. 120 120 100 _ 100 Ew Ew £6 =o 7° 3 4 aw 3 ° “200i 2002 2003 2004 2005 Sao leak ete 20012002 2003, 2008 2005 Fig. 1.18 Simple Line Chart Fig, 1.19 Line Chart with Symbolic Data Point ‘There are many form or variations of line chart of line graph, depends on the data points to be plot, for example step line chart, reverse step line chart, vertical segment line chart, horizontal segment line chart, curve line chart. (Vi) Area Chart — Area chart is also called area graph, use to display Quantitative data graphically. Area chart control is use represent data is bounded area. The bounded area is based on the line graph, the line is generated and the area below is shaded with colors, different texture and hatching, which produce area graph as shown in the figs. 1.20 and 1.21. Innation 100 80 oo 0 20 Inflation Rate 2001 2002 2003 2004 2008 2006 2007 2008, Fig. 1.20 Line Graph44 Machine Leaming (Vi-Sem.) nftation, 100 ‘0 « 40 20 Inflation Rate Soot 2002 2003 2004 2008 2006 2007 2008 Fig, 1.21 Area Chart ‘There are many form of area chart or line graph, depends on the data points to be plot, for example —step area chart, curve area chart while segmented | area chart is default area chart as shown in the fig. 1.22. 100 io fation Food Bi Inflation Overall nftation Rate $oor 2002 2003 2004 2005 2006 2007 2008 22 Multi Series Area Chart Re 3 oa se Probst. Find the probability of getting 4 heads in 6 tosses of afi c0% | 1 Sol. Here, p= 57 = 1—P= Le We know that Units) 45 afeGiel Gh RHC 160 60, 12, 7 mL ya 7 wl = 160+ 60+ 12 +1 =233 Ans. Prob.3. Alot has 10% defective items. Ten items are chosen randomly {rom this lot. Find the probability that exactly 2 of the chosen items are defective. (R.GPY, Dec. 2016) 2, n=10,r=2 Sol Here P 0 P= *capta* 2g 10-2 27 98 =e (LY( 2). ¢,f1) (2 7 lis) (ia) 7 ox) (3) 0x9 2x1 -(0.01)(0.4305) = 0.1937 Ans, yee Aperfect cubical die is thrown a large number of times in sets ie occurrence of S or 6 is called a success. In what, ion of the sets you expect 3 success. cree Ss. = 12 el Here, n=8, p= 3.q= 5 als ’ ‘Thus the Binomial distribution is given by N (344) ‘The number of sets in which 3 success are expected 37)8 =n ]8¢. (3) & on 56x32 { AG), 2x23 Percentage =N x 20232, 100 179200 : DDS * yeep BI Ans. "0b.S, The probabil the agUbS. The probability that a bomb dropped from a plane will strike “ US. If six bombs are dropped find the probability that S84 two will strike the target At Least two will strike the target. (R.GPL., June 2013)‘Sol. Given the probability of bomb that will strike the target is p a -4 5 “5 (})_ Therefore the probability of exactly 10 Will strike thet out of 6 bombs mm 6, pg? 2 744 6s (1)? (4)*__7680 =6, ptgt= SS /4) (4) - 7680 _ Capea “ (4) {3 31250 ~ 9246 An. (ii) The probability atleast two wall strike the target out of 6 bom = 1-@Probability that either none, or one are not strike the tage [fst] (2) 4.8) <1 tee 2] Ss) [5 5. 3125| 5 2048 107 =~ 395 * 3157 0345 Ans. Prob.6. The probability that an evening college student will graduate 0.4, Determine the probability that out of 5 students () None (ii) One and (iii) At least one will graduate. (R.GPY, Dec. 2015, 2019 ‘Sol. Given, the probability that an evening college student will graduat isp=04. ‘The probability that an evening college student will no graduate is q= 1-0 = 06. (i) The probability of none will graduate out of 5 student =5Cy po > 5Cq x (0.4)? * (0.6) - =1* 1 «0.07776 ~ 0.078 we ‘of exactly one student will graduate out of 55 $C, (0.4)! (0.6)* we 5x 0.4 x 0.1296 = 0.2592 wes (iii) The probability of at least one will graduate out of 55 = 1 (Probability of none will gadU0®) yy 10.078 = 0.922 a Unit 47 prob.7.1f 10% of bolts produced by a machine are defective, determine ‘pability that out of 12 bolts, chosen at random (i) one (i) none (ii) at the ig bots will be defective. ‘most v0 ‘Sol. Probability of defective bolts, p= Probability of not defective bolts, q ‘Total number of bolts, n= 12 (i) Probability of one defective bolt !2c,(0,1)!(a9)!" =12* 0.1 «03138 = 0.3766 Au. (ii) Probability that none is defective = Probability of 0 defective bolt = P(0) = cocoa)? 1x 1 * 0.2824 = 0.2824 Probability of two defective bolts 0% = 0.1 -0.1=09 Ans, xi 1x2) =%6,(a2(09)!9= 00103487 = 0.2301 Probability of at most two defective bolts P(O or 1 or 2) PCO) + PCI) + PQ) 2824 + 0.3766 + 0.2301 = 0.8891 Ans. Prob.8. An irregular six. ‘faced die is thrown and the expectation that in 10 ‘throws it will, sive five even numbers is twice the expectation that it will ‘ive four even numbers. How many times in 10,000 sets of 10 throws would You expect it to give no even number ? ‘Sol Let p be the probability of getting an even number. The probability of § even numbers in 10 throws 10C, p’ gS The Probability of 4 even numbers in 10 throws = !9C, p* qf low according to the given condition Cs pF gS=2 x MC, pt gs W976 gg yy NO9KBXT 4g Sx4x3x2x1 “" 4x3x2x1 8 > SL 5 P= 2q=> =(1-q)= q=>3-3q=5q S 8453 or gud Hence the ‘mes. nee the numberof times in 10,000 throws, where we getno even number 0,000 (3/8)! = 9.55 Ans.