International Scholarly Research Notices

Research Article

Online Boosting Algorithm Based on Two-Phase SVM Training

Algorithm 2

Proposed algorithm.
Input: Iteration numbers 𝑖 ,   𝑖 𝑤 , → 𝑥 , 𝑦 , 𝜆 and 𝜆 𝐻 , vectors → 𝛼 𝑘 ,
𝑘 = 1 … 𝑇 , → 𝛽
 for   𝑡 = 1 → 𝑇   do
   𝑓 𝑡 ( → 𝑥 ) = ⟨ → 𝛼 𝑡 , → 𝑥 ⟩
   ℎ 𝑡 = s i g n ( 𝑓 𝑡 ) ,
 end for
   → ℎ = ℎ 1 , … , ℎ 𝑡 , 1
   𝐹 = ⟨ → 𝛽 , → ℎ ⟩ ,
   𝐿 = m a x ( 0 , 1 − 𝑦 𝐹 ) ,
   𝑖 𝑤 = 𝑖 𝑤 + 1
  Pegasos ( → 𝛼 𝑇 , 𝜆 , → 𝑥 , 𝑦 , 𝑖 𝑤 , 𝐿 )
  Pegasos ( → 𝛽 , 𝜆 𝐻 , → ℎ , 𝑦 , 𝑖 , 1 )
 if   𝑖 𝑤 > 𝑟   then
   𝑖 𝑤 = 0
   𝑇 = 𝑇 + 1
   → 𝛼 𝑇 = → 0
   𝛽 𝑇 = 0
 end if