Multi-view　learning　based　on　graph　convolutional　networks　boosts　performance　by　incorporating　diverse　perspectives,　leading　to　significant　achievements　and　successful　applications　across　various　academic　and　practical　fields.　However,　multi-view　graph　convolutional　networks　suffer　from　substantial　computational　challenges　on　large-scale　graphs.　To　address　this　limitation,　graph　condensation　has　emerged　as　a　promising　direction　by　creating　a　smaller　composite　graph　that　allows　for　efficient　network　training　while　preserving　performance.　Furthermore,　previous　studies　have　demonstrated　that　encouraging　performance　in　graph　learning　is　achieved　via　graph　compression.　To　this　end,　we　attempt　to　introduce　graph　condensation　into　the　multi-view　learning　for　computation　acceleration.　This　approach　not　only　reduces　training　costs　significantly　but　also　achieves　sub-linear　time　complexity　and　memory　consumption　during　network　training.　Further,　we　propose　a　gradient　flow　induced　graph　convolutional　network　from　partial　differential　equations,　which　offers　theoretical　guarantees　and　potential　new　insights　for　the　graph-related　network　architecture　construction　with　transparent　model　interpretability.　Extensive　experiments　on　seven　real-world　multi-view　datasets　demonstrate　that　the　proposed　method　sharply　decreases　model　training　time　while　ensuring　competitive　multi-view　semi-supervised　classification.