Comments for Four Years Remaining

Comment on Everything New is a Well-forgotten Old by Dating

Dating — Mon, 11 May 2026 10:59:58 +0000

Dating Coach Amsterdam delivers transformational coaching
sessions. Dating Coach Amsterdam enables clients to facilitate confidence.
Dating Coach Amsterdam provides support on using dating apps efficiently.

The Dating Coach Amsterdam at Dating Coach Amsterdam delivers love and relationship
coaching. The Dating Coach Amsterdam from Dating Coach Amsterdam supports residents in managing emotions.
Dating Coach Amsterdam streamlines understanding of small talk in dating scenarios.
Dating Coach Amsterdam delivers clarity for clients seeking authentic relationships.
Dating Coach Amsterdam delivers coaching systems in The Netherlands.
The Dating Coach Amsterdam at Dating Coach Amsterdam allows
intimacy in dating. The Dating Coach Amsterdam from Dating Coach Amsterdam assists
meaningful dating experiences. Dating Coach Amsterdam provides guidance on developing confidence.
Dating Coach Amsterdam enables clients to navigate Amsterdam dating.
Dating Coach Amsterdam offers relationship coaching
for couples. The Dating Coach Amsterdam at Dating Coach Amsterdam delivers dating coach services.
The Dating Coach Amsterdam from Dating Coach Amsterdam helps residents
in achieving clarity. Dating Coach Amsterdam simplifies transformation through authentic dating experiences.

Dating Coach Amsterdam offers support to residents on managing emotions.

The Dating Coach Amsterdam at Dating Coach Amsterdam
provides coaching systems in Amsterdam. The Dating Coach Amsterdam from Dating Coach Amsterdam assists clients in facilitating confidence.
Dating Coach Amsterdam provides significant relationship coaching in The Netherlands.

Comment on The Data Science Workflow by Oli

Oli — Sun, 31 Jan 2021 12:13:12 +0000

Hi 🙂

so this is really an awesome post. I wonder, not more people commented :O

In the last month and years, the tools used for data science increased by alot. Is there any chance, that you can update your tool-list. That would be very interesting 🙂

Kind regards,
Oli

Comment on The Curse of Genomic Coordinates by Adam

Adam — Wed, 21 Oct 2020 05:01:43 +0000

Just another testimony that this was a useful script for me. I couldn't figure out how to make any vcf file remapping program to work so I used this library and strung together a simple python script and finally had success. Thanks!

Comment on How to Send an SMS by second

second — Mon, 05 Oct 2020 02:59:12 +0000

Hey there! I could have sworn I've been to this website before
but after reading through some of the post I realized it's new to me.
Anyways, I'm definitely glad I found it and I'll be book-marking and checking
back often!

Comment on The Mystery of Early Stopping by 6 Regularization Techniques for Deep Learning | Python | Keras - AI ASPIRANT

Mon, 03 Feb 2020 01:31:23 +0000

[…] Image SourceBy using the early stopping callback, which is available in Keras, we can monitor specific metrics like validation loss or accuracy. As soon as the chosen metric stops improving for a fixed number of epochs, we are going to stop the training. […]

Comment on What is the Covariance Matrix? by Aldo D'Ottavio

Aldo D'Ottavio — Fri, 13 Dec 2019 01:13:27 +0000

Very nice note! Thank you for taking the time to write it up.

Do you have a book or article reference which you would recommend for this?

Comment on ROC Area-Under-the-Curve Explained by Konstantin

Konstantin — Sun, 20 Oct 2019 21:10:41 +0000

The actual threshold value plays no role here. Any monotonous transform of the classifier's output will produce the same ROC curve with the same AUC score.

Note that as you move through the sorted list of classifier outputs you are effectively going through the various threshold values without using those values explicitly.

Comment on ROC Area-Under-the-Curve Explained by Subhrajit Bhattacharya

Subhrajit Bhattacharya — Mon, 30 Sep 2019 16:12:13 +0000

I liked your explanation.

Even though I do not understand the explanations using TPR/FPR fully (and I do not think it is worth understanding it fully), one thing I do understand is that the AUC is drawn by varying the threshold probability for declaring an example to belong to the positive class.

I do not think your explanation covers the role of the threshold value, no?

Cheers.

Comment on What is the Covariance Matrix? by Similarity, Distance/Metric, and Manifold Learning II | allenlu2007

Similarity, Distance/Metric, and Manifold Learning II | allenlu2007 — Mon, 19 Aug 2019 07:05:43 +0000

[…] [4] Konstantin, fouryears.eu, “What is the Covariance Matrix?“ […]

Comment on When the Best is not the Best by Konstantin

Konstantin — Sat, 04 May 2019 20:12:36 +0000

You reversed the numbers in your example, that's why it seems nonsensical to you. Here's how the correct example should look like: Suppose that no banks in the region in general offer loans cheaper than 6% to the majority of customers. Suppose that John is one of this majority and can expect a 6% APR loan. Knowing the market, John is happy take a 6% loan. Finally, suppose that bank X has somehow more data about John and a super-precise risk model, according to which John is super-trustworthy and even a 2% APR loan to him is still expected to be profitable. Does it make sense for bank X to give John a 2% loan when the bank knows it could just as well proudly issue a 5% loan and still make John happy with the "best rate"? Which model, in this case, would be better for the bank? The mathematically precise one, offering John a 2% loan, or a ROC-wise-worse one, which would "use the extra income" from John to finance a couple of more risky customers (which would then bring extra income in closing fees or add-on services)? The scrutiny of regulators cannot really prevent you from including ad-hoc criteria (such as "nicer-looking score spread") when you are training your scoring model.