?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Conditional+Variational+Diffusion+Models&rft.creator=della+Maggiora%2C+G&rft.creator=Croquevielle%2C+LA&rft.creator=Deshpande%2C+N&rft.creator=Horsley%2C+H&rft.creator=Heinis%2C+T&rft.creator=Yakimovich%2C+A&rft.description=Inverse+problems+aim+to+determine+parameters+from+observations%2C+a+crucial+task+in+engineering+and+science.+Lately%2C+generative+models%2C+especially+diffusion+models%2C+have+gained+popularity+in+this+area+for+their+ability+to+produce+realistic+solutions+and+their+good+mathematical+properties.+Despite+their+success%2C+an+important+drawback+of+diffusion+models+is+their+sensitivity+to+the+choice+of+variance+schedule%2C+which+controls+the+dynamics+of+the+diffusion+process.+Fine-tuning+this+schedule+for+specific+applications+is+crucial+but+time-consuming+and+does+not+guarantee+an+optimal+result.+We+propose+a+novel+approach+for+learning+the+schedule+as+part+of+the+training+process.+Our+method+supports+probabilistic+conditioning+on+data%2C+provides+high-quality+solutions%2C+and+is+flexible%2C+proving+able+to+adapt+to+different+applications+with+minimum+overhead.+This+approach+is+tested+in+two+unrelated+inverse+problems%3A+super-resolution+microscopy+and+quantitative+phase+imaging%2C+yielding+comparable+or+superior+results+to+previous+methods+and+fine-tuned+diffusion+models.+We+conclude+that+fine-tuning+the+schedule+by+experimentation+should+be+avoided+because+it+can+be+learned+during+training+in+a+stable+way+that+yields+better+results.&rft.subject=Denoising+Diffusion+Probabilistic+Models%2C+Inverse+Problems%2C+Generative+Models%2C+Super+Resolution%2C+Phase+Quantification%2C+Variational+Methods&rft.publisher=International+Conference+on+Learning+Representations+(ICLR)&rft.date=2024-01-16&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A++12th+International+Conference+on+Learning+Representations%2C+ICLR+2024.++++International+Conference+on+Learning+Representations+(ICLR)+(2024)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10198875%2F1%2F5967_Conditional_Variational_D.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10198875%2F&rft.rights=open