Open denoiser-ml opened 3 weeks ago
Hi! Thanks! Well, I think that the difference in performance comes mostly from the type of noise we are considering. In the case of the simulated process, we used some simulated noise with constant parameters throughout the simulation. However, in the real case since we don't know the nature of the noise, its characteristics could be variable or it can be noise from different nature in different sections of the sequences, etc... So probably we would need much more training data of the real process to let the network generalize well to this system and eliminate the noise better.
Best Regards.
From: denoiser-ml @.> Sent: Monday, June 10, 2024 4:45 AM To: salangarica/CBDAE @.> Cc: Saúl Alberto Langarica Chavira @.>; Mention @.> Subject: [salangarica/CBDAE] synthetic and real data noise (Issue #2)
Hi @salangaricahttps://github.com/salangarica , I have read your paper and I think it is a great work. I am studying your code and need more time to understand it better. At the moment I can't understand why your model performs so well with synthetic data but it is not very good with real data from sensors? Do we have different type of noise for the datasets you took in consideration? Kindly let me know. Thanks. Best regards.
— Reply to this email directly, view it on GitHubhttps://github.com/salangarica/CBDAE/issues/2, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABKW6PSAJYN7XYZTQRV6BTDZGVRSDAVCNFSM6AAAAABJB22YRKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM2DGMRYGEYDKMI. You are receiving this because you were mentioned.Message ID: @.***>
Hi @salangarica , I have read your paper and I think it is a great work. I am studying your code and need more time to understand it better. At the moment I can't understand why your model performs so well with synthetic data but it is not very good with real data from sensors? Do we have different type of noise for the datasets you took in consideration? Kindly let me know. Thanks. Best regards.