林彦
2018-09-09
均值函数不为0就会影响高斯过程产生的曲线的分布。比如一个线性上升的均值函数对应的多元高斯分布曲线会有线性上升趋势。
作者回复: 简单来说，直接使用零均值会让置信区间越拉越大，预测的不确定性也就越来越大。随着测试数据离训练数据越来越远，预测结果的波动会逐渐变得强烈。
<div class="router-button__Wrapper-sc-1xlmfkg-0 jDKtMA"><div class="frame-button-content__Wrapper-sc-1td9tx7-0 huSWDo"><div class="frame-button-content__BottomLeft-sc-1td9tx7-1 dmMDDL"><label class="icon__Span-sc-5foskt-0 cSUETK"><span type="chat-button-1" width="32px" height="32px" class="base-icon__Container-sc-1ajxhaa-0 ihFnZp"></span></label></div><div class="frame-button-content__ButtonRight-sc-1td9tx7-2 cSpVKy">我们在线，来聊聊吧</div></div></div>
<div class="bubble-arrow__Arrow-sc-3uxf0p-0 bfDInF"></div><div class="router-bubble__Wrapper-k6dy4k-0 itiunK"><div class="frame-bubble-content__Wrapper-lbgrtb-0 diLqpO"></div></div>
<div class="router-invite__Wrapper-sc-1m23l53-0 gORLfI"><div><div class="frame-invite-content__CloseWrapper-sc-1g53vrw-2 eApwIP">✕</div><div class="frame-invite-content__FrameInviteContentBoxStyle-sc-1g53vrw-0 AkiIQ" transform="translate(20px,20px)"><div class="frame-invite-content__FrameInviteContentStyle-sc-1g53vrw-1 blgYRf" width="260px" height="44px"><div class="content">您好，当前有专业客服人员在线，让我们来帮助您吧。</div></div></div></div></div>
<div class="router-call-botton__Wrapper-dkoslj-0 fkxFou"><div class="frame-button-content__Wrapper-sc-1td9tx7-0 fxjwak"><div class="frame-button-content__RoundWrapper-sc-1td9tx7-3 LOaHP"><label class="icon__Span-sc-5foskt-0 cSUETK"><span type="call-callback" width="32px" height="32px" class="base-icon__Container-sc-1ajxhaa-0 cvRvuJ"></span></label></div></div></div>
我们在线，来聊聊吧
2018-09-09均值函数不为0就会影响高斯过程产生的曲线的分布。比如一个线性上升的均值函数对应的多元高斯分布曲线会有线性上升趋势。