Download as pdf
Download as pdf
You are on page 1of 2
: For which ofthe following tasks might means clustering be a suitable algorithm? Solect al pont that apply. [chen many mats, you wantto determine they ere Spam or Nor Spam ens (BH cvenasetofnews artes om many citferen news ebsites, And aut wh are te main opis evered, [71 Gren historical weather records, predict if tomorrows weather willbe sunny oF rainy, From the user usage patterns on a website, igure out what different groups of cnmmmatnsne my] -[] etn Ef Furmerore, we ave orang exampie 0 — [2]. aera duster asgnmerstep, hat wi be? O Oe Ol isnot assigned oy 1 3. k-means is an iterative algorithm, and two of the following steps are repeatedly carried out in point Its innerlaop. Which two? “The cluster centroid assignment step, where each cluster centroid pis assigned (by setting eto the closest waning example 209 “The cluster assignment step, where the parameters € are updates. Move each duster centroid fg, by setting it to be equal tothe closest training example 2) TE) stove the cluster centroids, where the centro ny are update. | 4s. sueposeyou have an unable tas {2 wn tandore 21}, You run Kemeans with S0 different {nialzations, and obtain 50 afferent clusterings of the data. What isthe recommended way for choosing which one of ‘these 50 dusterings tose? Use the elbow method. Manually examine the lusterngs, and pick the best one. © Plocthe data and the cluster centro, and pick the clustering that gues the mast ‘coherent cluster centroid, Compute the distortion function J(et) {at minimizes this 60), yy dg). and pick the one + | 5. Whichot the folowing staterents ae true? Select al hat apply. seit | te standord way of ntiaing Kmeons is seting py vector ofzeros my tobe equal toa since kadeansis an unsupervised erring algartn itcannc overt the dt, nd crus aay beter to hae a ge a umber of str ass computation teasble [BE tre are woried about means geting suckin bad local optima, one way to amelie ede) this prob is if we ty using mip random inaliatons (BE Forsome datasets the “right or“orec valve afk the number of clusters canbe ambiguous anchard even for ahumanexpertooking careful athe data to cecige

You might also like