Local Beam Search (Classical)

Start the search from K randomly selected states.

S1 S2 . . . . Sk-1 Sk
b Repeat with k
Branching
factor is b best states from
the pool

T1 T2 . . . . Tc-1 Tc
40 10 60 90

Pool of C states, generated from initial k states… Where, c = b x k

Search Landscape
 Stochastic Local Beam Search Repeat with k
Start the search from K randomly selected states. states based
on selection
probability Pi
S1 S2 . . . . Sk-1 Sk
b
Branching
factor is b

T1 T2 . . . . Tc-1 Tc
40 10 60 90

Pool of C states, generated from initial k states… Where, c = b x k

Selection Probability of ith State:
𝐹𝑖𝑡𝑛𝑒𝑠𝑠 (𝑇𝑖 )
𝑃𝑖 = σ𝑐𝑖=1 𝐹𝑖𝑡𝑛𝑒𝑠𝑠 (𝑇𝑖 )