Intelligent　reflecting　surfaces　(IRS)　can　dynamically　change　the　wireless　channel　and　improve　the　communication　performance　by　subtly　adjusting　the　signal　reflection　through　a　large　number　of　low-cost　passive　reflecting　elements,　which　has　become　the　focus　of　wireless　communication　research.　However,　due　to　the　addition　of　IRS,　the　whole　communication　system　becomes　more　complex　and　the　dynamics　of　the　system　is　higher,　making　the　communication　system　face　many　new　challenges.　With　a　strong　data　processing　and　adaptive　ability,　machine　learning　(ML)　can　constantly　adapt　to　changing　environments　and　needs　and　well　cope　with　many　challenges　in　communication　systems.　Therefore,　employing　ML　to　tackle　issues　in　IRS-assisted　communication　systems　has　emerged　as　a　central　theme　of　current　research.　Based　on　this,　this　paper　gives　a　systematic　overview　of　the　application　of　ML　in　IRS　system.　Starting　with　the　problems　existing　in　IRS-assisted　communication　system,　this　paper　expounds　and　analyzes　the　application　of　ML　in　IRS　system　from　four　aspects:　design　and　optimization　of　reflection　factor,　channel　processing　and　modeling,　resource　allocation　and　management,　and　security　enhancement,　and　it　also　discusses　the　advantages　and　future　development　trends　and　challenges　of　applying　ML　in　IRS　system.　©　2025　Beijing　University　of　Technology.　All　rights　reserved.