Sergii Kharlanov

Category: ML

  • STDs can be different

    —

    by

    sergii_kharlanov
    in

    std – article Innocent beginning¶ I was figuring out normalization and came across some code where the MNIST dataset was divided by 255 before normalization. I began to suspect, based on observations from this article: How ML models work with images? that this is incorrect. Not incorrect, but unnecessary, because normalization should “handle” this division…

  • How ML models work with images?

    —

    by

    sergii_kharlanov
    in

    load_images_for_ML Intro¶ If you read chapter 4 from Practical Deep Learning for Coders and you have questions about all these numbers from images like what is $28*28$ or $784$ or why we divide these tensors with image data on 255 then this is a post for you Usual start of working with images¶ In [1]:…

  • RMSE / L2 normalization

    —

    by

    sergii_kharlanov
    in

    This is loss metric for machine learning models.Has two names: Computing RMSE To compute RMSE we need to Actual Predicted Error Powered error 10 20 -10 100 3 8 -5 25 6 1 5 25 Total powered error: 150Mean powered error: 50Root mean squared error: 7.07 Python code to compute RMSE – SciKit Python code…

  • MAE / L1 normalization

    —

    by

    sergii_kharlanov
    in

    This is loss metric for machine learning models.It has different names: Computing MAE To compute MAE we need to: Actual Predicted Error 10 20 -10 23 13 10 5 6 -1 Total absolute error: 21Mean absolute error: 7 Why absolute values? If we compute non absolute difference, then negative errors can cancel out positive errors:…