Such a clever new LM architecture: Orchid uses data-dependent convolutions to overcome the quadratic complexity of traditional attention. 🌺 Achieves quasilinear scalability with dynamic kernel adjustments for long sequences! Also, super strong as a BERT model!
Paper: https://t.co/U4ImzVHsia
No code yet, but trying to implement it (as a pLM, of course) before ICLR. 👨🏾💻
Mohammadreza Fani Sani @fanisanimis the first PhD graduate of the @pads_rwth chair!
Last week he successfully defended his thesis, "#Preprocessing Event Data in #ProcessMining".
Ad Maiora! Best of luck for your career!
📰 Can't leave you without anything to read on the weekend!
New paper accepted in the Journal of Intelligent #InformationSystems#JIIS: "Performance-Preserving Event Log #Sampling for #PredictiveMonitoring". Full text available at https://t.co/DHfgOqaVWp
Author list below 👇
وصف این روزهای #دانشگاه_شریف: تحمیل دیدگاه اقلیت بر اکثریت.
مثلاً هفتهی قبل در جلسهی استادان با هیات رییسه، «همهی» استادانی که صحبت کردند بر حذف ممنوعالورود شدن دانشجویان تاکید کردند.
در نتیجه از امروز وارسی کیف دانشجویان (طبق اطلاع من برای نخستین بار در تاریخ شریف) آغاز شد.
I have a student who has a problem drawing by hand due to some disability. Wondering if there is any speech-to-model (business process or general) modeling tools to help? such help sounds to be more effective than personal assistant #BPM#BPMN#modeling#teaching