The　personalized　recommendation　has　already　taken　a　crucial　role　in　online　services　to　alleviate　information　overload.　However,　most　existing　works　pay　their　attention　to　user　interest　modeling　with　a　uniform　embedding,　which　inevitably　results　in　suboptimal　recommendations.　We　argue　that　users＇　diverse　and　mixed　interests　are　positively　related　to　their　interacted　items　with　mutual　effects,　and　they　are　partially　matched　in　decision-making.　This　work　introduces　a　novel　dual　interest　activation　network　(DIAN),　which　forms　dual　embedding　learning　of　users　and　items　with　a　dynamic　activation　for　personalized　recommendation.　By　modeling　mutual　effects　between　users　and　items,　DIAN　constructs　dual　embedding　learning　on　their　identites　(IDs)　and　neighbor　groups　to　encode　their　diverse　and　mixed　composition　of　interests　implicitly.　Specifically,　considering　the　partial　matching　scenario,　we　introduce　a　dynamic　interest　activation　with　pairwise　matching　motivated　attention　on　aggregating　neighbor　groups　of　users　and　items.　Experimental　analysis　verifies　the　significance　of　mutual　effects　and　dynamic　matching,　illustrating　the　effectiveness　of　DIAN　for　personalized　recommendation.