In　most　of　the　existing　multi-view　subspace　clustering　methods,　the　consistent　shared　information　of　the　multi-view　data　is　learned,　and　the　contribution　of　each　view　is　regarded　as　equally　important　to　integrate　the　difference　information　of　multiple　views.　However,　possible　noise　or　redundancy　between　different　views　is　ignored　due　to　the　idea　of　treating　each　view　as　equally　important,　resulting　in　poor　final　clustering　performance.　Therefore,　an　algorithm　of　adaptive　weighted　multi-view　subspace　clustering　guided　by　manifold　regularization(MR-AWMSC)　is　proposed　in　this　paper.　The　consistent　global　low-rank　representation　information　for　each　view　is　learned　by　nuclear　norm,　and　difference　information　from　different　views　is　described　by　group　effect.　According　to　the　concept　of　manifold　regularization,　the　weight　of　each　view　is　adaptively　learned,　and　the　contribution　degree　to　the　difference　information　of　each　view　is　automatically　assigned.　The　difference　information　is　integrated　by　the　adaptive　weight　and　the　consistent　information　is　fused　to　obtain　the　final　consensus　representation.　The　consensus　representation　is　constructed　for　clustering　multi-view　data.　Experimental　results　on　six　public　datasets　demonstrate　that　MR-AWMSC　effectively　improves　the　multi-view　clustering　performance.　©　2022　Journal　of　Pattern　Recognition　and　Artificial　Intelligence.　All　rights　reserved.