本稿では数値計算の際に生じる誤差（情報落ち）について、これを軽減するための関数を作成し、誤差の低減にどの位寄与するか調査を行った。 絶対値の大きな値と小さな値を足し合わせる時に、時々、計算機の有効桁数を超えるような演算が行われる。 情報落ちとは、有効桁数の範囲外となって切り捨てられた値の蓄積により生じる誤差の事で、 数値積分などで棒大な数の短冊状の領域を足し合わせる場合等で注意しなければならない。

In this report, I examined the influence of computation error (information loss). The content is making of summation function for information loss measure and compared non-measured calculation result and measured one. In summation between large absolute value and small one, sometimes, unfavorable calculations which surpasses effective digit are carried out on computers. Here, "information loss" means a computation error caused by accumulation of values which became outside of the effective digit and were cut off. For example, we have to calculate carefully when calculate numerical integration, like summation of many small value.