Arguments

Map Reduce

  1. MR steps description
  2. MR performance analysis tools
    every sequential con be done in 1 step MR - how and why is bad
    design goals of MR
    pros and cons
    what is partitioning
    class count problem (description)
    value of l for best tradeoff when partitioning
    how large m can be? (proof)

Corset Technique

  1. what is corset techique and composable corset technique
    metric space
    distance functions (Minkwski, Angular, Hamming, Jaccard)
    aims of k-center, k-means, k-median
    FFT (what + pseudo)
    FFT 2 approximation (proof)
    MR FFT (how + analysis)
    Lemma (proof)
    MR FFT 4 approximation ((proof)
    Observation on low-dimensional pointsets for MR FFT
    worst diameter 2 approximation (proof)
    better coreset-based diameter approximation (proof)
    diversity maximization (definition)
    (1+e) coreset for div max
    relation between div max and k-center problem
    how coreset-base algo for div max ((proof)
    Lloyd's algo, k-means++, PAM (just main idea)
    coreset-based for kmeans-kmedian
    weighted k-means + algorithm variation
    MR-kmeans(A) (general idea + analysis + accuracy)

Streaming

streaming model (describe)
performance metric in streaming
majority problem
Boyer-Moore algorithm (metrics + correctness)
sampling
sampling in streaming
reservoir sampling (pseudo + analysis)(proof)
frequent items problem
e-AFI problem
sticky sampling (pseudo + analysis)(proof)
frequency moments
probabilistic counting algorithm (def + prob guarantees)
count-min sketch (ingredients + pseudo + th + prof --> kinda skip)
count sketch (ingredients + pseudo + th + prof)
performance metrics of count-min and count sketch
approximate membership problem
bloom filter (ingredients + pseudo + false positive th + prof)

r-NNS problem (definition)
Other variations (just the difference)
kd-tree (what is)
RR problem (definition)
Performance of kd-tree in R^2 
Generalized performance in R^D
(c,r)-ANNS (definition)
(p1,p2,c,r)-LSH (definition)
ANNS with LSH performance (+ prof)