Problems in the current research on forecasting with transformers, foundational models, etc.

Поделиться
HTML-код
  • Опубликовано: 23 дек 2024

Комментарии • 8

  • @bobsoup2319
    @bobsoup2319 День назад +3

    Had me captivated the whole time without any fancy graphics or special production. This speaks to the amazing quality and quantity of the information you shared

  • @AhmedThahir2002
    @AhmedThahir2002 2 дня назад +1

    Fantastic talk!!
    I recently did a similar project for my undergrad highlighting how Machine Learning models for predicting inflation rate, and basic linear regression and naive benchmark were winners - I thereby questioned the results of a high profile paper claiming ML models' success.
    Unfortunately, one of the evaluating professors was not very happy and just asked if I did it right. When I tried to explain how I ensured my methodology, he interrupted me and asked me to revisit my project. I totally respected the fact that he asked me to recheck my methodology, but not allowing me to explain how I disproved a top paper in 5 months was a bit upsetting.
    Talk about p-hacking and survivorship bias in academia!

  • @4thpdespanolo
    @4thpdespanolo 2 дня назад +1

    Great talk. I’m now inspired to publish meta analyses on foundational model results from benchmark datasets…

  • @JohnDoe-sy8bi
    @JohnDoe-sy8bi День назад

    Have you looked into temporal fushion transformers? those at least promise interpretability

  • @GabrielFuentes-u4q
    @GabrielFuentes-u4q 2 дня назад +2

    amazing stuff! thank you very much for sharing

  • @mailailuan
    @mailailuan 2 дня назад +1

    Great talk!

  • @oliverzhang9839
    @oliverzhang9839 День назад

    greak talk!❤