Mikael Huss

@mikaelhuss

Data person, blogger , active in Sthlm R useR group and .

Stockholm
Vrijeme pridruživanja: travanj 2007.

Tweetovi

Blokirali ste korisnika/cu @mikaelhuss

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @mikaelhuss

  1. proslijedio/la je Tweet
    31. sij
    Poništi
  2. proslijedio/la je Tweet
    28. sij

    Ever wonder, are there any machine learning applications actually being used to care for patients in health care? If yes, check out our new review: In it, we ( ) present 21 ML products translated into clinical care THREAD

    Prikaži ovu nit
    Poništi
  3. 27. sij

    I've blogged about Google's sequential attention based method TabNet for prediction on tabular data . Here is a link to the rudimentary CLI I built to make it more convenient to train models with it

    Poništi
  4. 20. sij

    I've blogged about my go-to tool for prediction on tabular data, CatBoost, and its differentiable (neural network based) counterpart, NODE. Future blog posts on other tools to follow!

    Poništi
  5. proslijedio/la je Tweet
    12. sij

    The funny thing about AIDungeon is due to all the GPT-2 pre-training on swaths of random web text, you can create an adventure game about literally anything. Here's one about installing OpenCV from source:

    Poništi
  6. proslijedio/la je Tweet
    12. sij

    Our variable importance paper is out in JMLR!! Often, many prediction models achieve high accuracy. A variable could be important to some models, but not others. We study how important a variable could be to *any* model (in a pre-specified class). (1/6)

    Prikaži ovu nit
    Poništi
  7. 12. sij

    (The past few years are not included though; it's from 1963 to 2015)

    Prikaži ovu nit
    Poništi
  8. 12. sij

    Current temperature in Stockholm: higher than any temperature measured at noon on this date in 60 years. 8.3C vs previous record 7.3C from 1983.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    11. sij

    A computer scientist invented some new big data prediction techniques. Did a stint in industry. Warned about the potential for algorithmic discrimination and data-driven monopoly power. Sounds very 2020s. But it's Charles Babbage, in the 1820s.

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    2. sij

    Brains are amazing. Our lab demonstrates that single human layer 2/3 neurons can compute the XOR operation. Never seen before in any neuron in any other species. Out now in . Congrats Albert, Tim  & CO

    Poništi
  11. proslijedio/la je Tweet
    9. pro 2019.

    We used our "DNA of things" architecture to store the 3D printing file (stl file) of the bunny in DNA. Then, we took the DNA, encapsulated it in silica beads, mixed the beads with PCL plastic, extruded, and printed the bunny. The end result?

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet

    Like the classic competitions that accelerated the era of deep learning, the Human Protein Atlas competition had 2,171 team compete for analyzing ~78,000 images--> high performance models and collaborators

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet
    30. stu 2019.

    🥳 shinyMobile officially released on CRAN v0.1.0. Thanks CRAN maintainers for being so fast! Thanks and for their help 🙏. Have fun to develop shiny apps on and , with fullscreen capabilities

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    28. stu 2019.

    Some more information about the M5 Competition. It will start on the 2nd of March, 2020 and end the 30th of June 2020 It will be run using the Platform. There will be about 100K time series of sales data made generously available by as shown in the attached Table

    Poništi
  15. proslijedio/la je Tweet
    29. stu 2019.

    Can you increase ML model performance by removing data? I wrote this quick blog post on using data valuation to get better accuracy with less data.

    Poništi
  16. proslijedio/la je Tweet
    28. stu 2019.

    Memories are just dimensionality reduction of the past.

    Poništi
  17. proslijedio/la je Tweet
    25. stu 2019.

    Always test genomics methods with real data! This is a self-self dot plot.

    Poništi
  18. proslijedio/la je Tweet
    4. stu 2019.

    Incredible paper from and : Ultrafast search of all deposited bacterial and viral genomic data "entire set of 469,654 bacterial, viral and parasitic WGS datasets in the ENA" ... and ~65% were from 5 genera.

    Poništi
  19. proslijedio/la je Tweet
    9. stu 2019.

    Simple but effective trick: put your data on a log scale. A lot of real-world data is log-normal and changing the scale will help your predictions and visualizations

    Poništi
  20. 5. stu 2019.

    The RIKEN-KI-SciLifeLab symposium on biomedical data for artificial intelligence is kicking off in Yokohama!

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·