You are on page 1of 1

Summary of important definitions and formulas

Error definitions
True error
𝐸𝑡 = 𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒
True fractional relative error
𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒
𝜀𝑡 =
𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒
True fractional relative error in percent
𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒
𝜀𝑡 = ∗ 100%
𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒
Approximate relative error (can also be expressed in %)
𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑜𝑙𝑑 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒
𝜀𝑎 =
𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒
Critical tolerance error which ensures correctness to at least n significant figures
, n=number of significant figures𝜀𝑠 = (0.5 ∗ 10(2−𝑛) )%
We need |𝜀𝑎 | < 𝜀𝑠 𝑜𝑟 |𝜀𝑡 | < 𝜀𝑠 depending on the problem
Taylor series
𝒇(𝒙) = 𝒇(𝒂) + 𝒇́(𝒂)(𝒙 − 𝒂) + 𝒇(𝒂)(𝒙 − 𝒂)
First order derivatives
Forward Difference
𝒇(𝒙𝒊+𝟏 ) − 𝒇(𝒙𝒊 )
𝒇′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉)
𝒉
Backward Difference
𝒇(𝒙𝒊 ) − 𝒇(𝒙𝒊−𝟏 )
𝒇′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉)
𝒉
Central Difference
𝒇(𝒙𝒊+𝟏 ) − 𝒇(𝒙𝒊−𝟏 )
𝒇′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉𝟐 )
𝟐𝒉
Second order derivatives
Forward Difference
𝒇(𝒙𝒊+𝟐 ) − 𝟐𝒇(𝒙𝒊+𝟏 ) + 𝒇(𝒙𝒊 )
𝒇′′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉)
𝒉𝟐
Backward Difference
𝒇(𝒙𝒊 ) − 𝟐𝒇(𝒙𝒊−𝟏 ) + 𝒇(𝒙𝒊−𝟐 )
𝒇′′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉)
𝒉𝟐 Newton Raphson method for
Central Difference
𝒇(𝒙𝒊+𝟏 ) − 𝟐𝒇(𝒙𝒊 ) + 𝒇(𝒙𝒊−𝟏 )
solving two nonlinear equations
𝒇′′(𝒙𝒊 ) ≅ , 𝑹𝒏 = 𝑶(𝒉𝟐 )
𝒉𝟐 u(x,y)=0 and v(x,y)=0
Error propagation and condition number
∆𝑓(𝑥̃) = |𝑓′(𝑥̃)|∆𝑥̃
𝑓 ′ (𝑥̃)𝑥̃ 𝜕𝑣 𝜕𝑢
𝐶𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 𝑛𝑢𝑚𝑏𝑒𝑟 = | | 𝑢𝑖 𝜕𝑦𝑖 −𝑣𝑖 𝜕𝑦𝑖
𝑓(𝑥̃) 𝑥𝑖+1 = 𝑥𝑖 −
Newton Raphson 𝐽
𝒇(𝒙)
𝒙𝒊+𝟏 = 𝒙𝒊 − 𝜕𝑢 𝜕𝑣
𝒇′(𝒙𝒊 ) 𝑣𝑖 𝜕𝑥𝑖 −𝑢𝑖 𝜕𝑥𝑖
𝑦𝑖+1 = 𝑦𝑖 − 𝐽
Secant method
𝑓(𝑥 )(𝑥𝑖−1 −𝑥𝑖 )
𝑥𝑖+1 = 𝑥𝑖 − 𝑓(𝑥𝑖 )−𝑓(𝑥 ) 𝜕𝑢𝑖 𝜕𝑣𝑖 𝜕𝑢𝑖 𝜕𝑣𝑖
𝑖−1 𝑖
𝐽= −
𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥

You might also like