Tag: DataScience

Exploring ASML’s Advanced Chip-Making Equipment with Veritasium | January 02 2026, 00:47

Veritasium released a very cool report yesterday from ASML about the equipment used to print chips for your little phones, cameras, and laptops.

For those who aren’t familiar with the process. First, a monocrystal is grown from ultra-pure silicon and cut into thin wafers, then multiple layers of thin dielectrics, conductors, and semiconductors are repeatedly applied to the wafer surface, each time shaping the necessary areas using photolithography, etching, and ion doping, eventually creating billions of transistors and connecting metallic paths; finally, the wafer is tested, cut into individual crystals, and packaged into casings, making them into finished microchips.

This process had a limitation – the width of the paths and the distance to the next one are limited by the wavelength of the light used, and reducing it is difficult because there’s nothing to focus such a beam with – lenses simply absorb/reflect everything. In EUV lithography (extreme ultraviolet), the wavelength is 13.5 nm. This is virtually soft X-ray radiation.

The video explains details about the ASML machine costing 400 million dollars. Instead of refracting lenses, highly complex systems of reflecting mirrors are used. These mirrors are the smoothest surfaces ever created by humanity. If the mirror of this machine were enlarged to the size of the Earth, the largest bump on it would not be thicker than a playing card. To enable the mirrors to reflect X-rays, up to 76 alternating layers of tungsten and carbon, each less than a nanometer thick, are applied. All this is done by Zeiss. In addition, this mirror has a controlled curvature—it is constantly adjusted by robots with precision up to picoradians. The precision of the mirror control is so high that if a laser were mounted on it, directed at the Moon, the system could choose on which exact side of a 10-cent coin lying on the moon’s surface to hit with the beam.

But. We don’t have a “light bulb” that emits light in the EUV range.

To generate this light, a laser “shoots” at a droplet of molten tin the size of a white blood cell, traveling at 250 km/h. The first pulse flattens the droplet into a disc, the second and third turn this “disc” into plasma – and all this occurs within just 20 microseconds. When hit by the laser, the droplet heats up to 220,000 Kelvin — approximately 40 times hotter than the surface of the Sun. This plasma emits that very necessary light. And it does so 50,000 times a second. They say it’s been brought up to 100,000. Imagine, at a hundred thousand laser shots per second, it never misses a single one. All this happens in a deep vacuum. To clean the mirrors from tin particles, the chamber is constantly blown with hydrogen at a speed of 360 km/h — faster than a Category 5 hurricane. This process is described by the same formula (Taylor-von Neumann) that describes a nuclear explosion or supernova explosion.

The machine layers the chip with an error margin of no more than five atoms, while the matrix swings back and forth with an overload of 20G.

A single High-NA machine is transported in 250 containers on 25 trucks and seven Boeing 747 aircraft.

Link to the video – in the comments. Or search on YouTube on the channel veritasium.

Visualizing Volleyball Plays: A Glimpse into My App’s Functionality | January 01 2026, 21:37

Here is actually a quick screen capture of how my app for creating and visualizing volleyball schemes works.

Implementation details here: https://www.facebook.com/raufaliev/posts/pfbid0njrqH8oLcWGFsZcgE5o2pj3NcDNaYSQeCMY6twXNbZn6dc38m9kjhsBqA4YsMozcl

Decoding the Beast: Migrating from Excel to Code | December 17 2025, 18:56

We’ve all encountered it — the “Main Excel Spreadsheet Managing the Business.” The very one B2B companies use to calculate million dollar quotes. It has 12 tabs, 1000+ nested formulas, and zero documentation. For ten years, it had “quick fixes” slapped on and constants hidden away. It’s no longer just a file, but a living organism that no one fully understands except for the guy who quit years ago. That’s how puzzled I was. Moreover, there was uncertainty whether even half of the formulas were needed, or if they were vestiges of the past.

Typical cell:

=IF($D11=$D10,””, IF(ISNUMBER( INDEX(Data!$T$10:$U$17,

MATCH(TabCalc!$F11,Data!$T$10:$T$17,0),2)),

INDEX(Data!$T$10:$U$17, MATCH(TabCalc!$F11,Data!$T$10:$T$17,0),2),

INDEX(TabProd!$C$8:$U$112,TabCalc!$D11,I$1)))

I was tasked with transferring this logic into code so that it was all computed by software. The Excel file seemed to have everything it needed, but in reality — it was a complicated black box. 1069 formulas.

The challenge was in how to translate a thousand interdependent formulas into clean code without losing any edge cases.

Here’s what I ended up doing.

Instead of rewriting everything from scratch at once with uncertain prospects of bug proliferation, I used a strategy of lazy computations and mocks.

I built a structure on Groovy that mimicked Excel’s behavior. Each computation (from a cell) I defined as a function that executed only when it was called. And the functions were a multidimensional dictionary.

I started from the end of the computation graph: from results to inputs. If a formula depended on something I hadn’t yet written, I “mocked” it in the code, simply substituting the value from the Excel sheet.

Bit by bit I replaced these mocks with real logic. Comparing the output of my code to the Excel at each step, I could clearly see where my logic diverged.

In other words, I moved from the result to the input data. At each step, it was clear which mocks needed to be turned into code, and I could compare version +1 with version -1 — the result had to match. As soon as all mocks were replaced with calls — the task was done.

The real “secret ingredient” was the dynamic nature of Groovy for creating a multidimensional map of functions. Instead of static variables, I used a deeply nested structure, where each “leaf” was a closure. This allowed access to any part of the table — be it an input parameter, a config constant, or a complex intermediate result — through a simple, unified syntax, and some components were dynamic.

Here’s an example:

conf[“group”] = { x -> [“a”, “b”, “c”] }

conf[“group”]().each {

calculate[“Group”][“Subgroup”][it][“TotalQuantity”] =

{

x -> calculate[“Group”][“Subgroup”][it][“Someparameter”]() * conf[“someConstant”]()

}

}

Using dynamic keys and closures, I could iterate through product groups or data sets. Since these were dynamic functions, not stored values, the entire system worked like a living graph of dependencies.

Testing was possible right from the start of transferring the formulas. The charm was that you were kind of addressing a cell through syntax like calculate[“Totals”][“A”](), but in reality, you were launching an entire tree of calculations at that moment. And this was incredibly convenient for debugging.

In two weeks, the “Black Box” was transformed into a transparent, modular library with clear logic, which produced exactly the same result as the original table.

P.S. Of course, all the data in all the screenshots are thoroughly obfuscated, or rather, written from scratch for this text.

From Idea to Chess AI: Building a Neural Network to Predict Moves | December 15 2025, 04:33

While figuring out neural networks, I decided to come up with a game-related task for myself. What if I find some ready-made games, and train a neural net to predict moves based on the board situation. Said and done. Of course, generating code is faster with LLM, but I wrote the detailed assignment myself and designed the architecture on my own. In 40 minutes (!) from the idea to the result, I already had a working solution that, at least in the first half of the game, does not mess up too much.

In the screenshot is CuteChess – it works with any chess engine, and in my case, it’s a simple Python script. The script takes the board situation and feeds it to the model. It selects the top 5 moves, and only these top 5 are analyzed deeply for several moves ahead and assesses the position. That is, the neural network suggests possible moves based on the analysis of 20,000 games (534,453 positions). From the results, the best is chosen. It uses the minimax algorithm for this, if that means anything to anyone (it didn’t to me, so Gemini here helped me)

How the model is trained. On the lichess website, you can download games, there are hundreds of gigabytes. I took a file with 800,000 played games from the year 2014. From these 800,000, I select 20,000, specifically looking with a script for games where the result is not a draw (1-0 or 0-1). Next, I calculate the difference (Winner_Rating minus Loser_Rating). It’s not the best metric, but it’s better than nothing. The bigger this difference, the more “confident” the win should be (the strong punish the weak). Thus, I get 20,000 such games.

“Ignoring the moves of the weak” (to avoid teaching the model bad play) is implemented during the training stage of the model. Essentially, the logic is: “If it’s White’s turn now, and White won this game — we learn. If it’s Black’s turn now, and Black lost — we skip and don’t teach the net this move.”.

The neural network is trained in batches of 128 positions at a time. The network receives a board position as input and outputs 4096 — the probability assessment for each possible move.

Selecting games takes about 5 minutes. Training the model on my computer takes about 10 minutes for 20,000 games. You could leave it to train on 100K or a million, and it would definitely be better. No need anymore – I figured it out 🙂

You can view the game here:

https://lichess.org/JWeaIrVW

Exploring the Magic of Neural Networks in Letter Prediction and Visualization | December 14 2025, 23:35

I am currently experimenting with training simple neural networks – primarily to automate the existing toolkit, and some things just seem like magic.

There is a database of 32,000 names. There is a neural network filled with random numbers. I start training, with only this list of names as input. The first layer of the neural network is embeddings, and I set the number of dimensions to 2 for easy visualization. And after 200,000 iterations of training, the system clearly separates vowels from consonants, and for some reason, places the letter “q” slightly apart from other consonants. It seems that this is because the letter ‘q’ almost exclusively predicts the letter ‘u’ (Queen, Quincy, Quentin).

It also very reliably separates vowels and consonants in Russian names. In Russian names, the letters b and l are somewhat away from the other consonants, as are the soft and hard signs (well, that’s understandable).

I wonder how it works. If trained on a normal corpus of texts, the difference would be very clear. Why are vowels separated from consonants? Apparently, from the network’s mathematical perspective, ‘a’ and ‘o’ serve the same function: they “trigger” the prediction of the consonant following them, so the alternation of vowels and consonants is to blame. But damn, it’s interesting 🙂

And since the model can predict the next letters, you might try running it on Russian. On a model with 30-dimensional embeddings, it invents names like: Byaketta, Afsena, Erakey, Zasbat, Daraya, Gaiomahad, Rain, Razhul, Gzhatsiy, Reben, Vureb, Durodira, Turuzhul, Regravgava, Razsan, Gabila, Avganzh, Raksi, Khalebkokhorta, Rather. The model – for those who understand – is this: input of 6×33 characters (because we take up to 6 characters of context), encoded into embeddings of 60, goes to a layer of 100 neurons, and from there back to 33 characters. Some nonsense, but at least it’s clear how it all works at all levels.

Harnessing GPU Power Beyond Machine Learning: A Data Processing Experiment | December 13 2025, 01:16

Torturing my supercomputer. Illustration that the GPU is not just for machine learning and some complex math.

My script takes a thick English dictionary (Webster) and multiplies it by 30, creating a list of 12 million words. Then, the algorithm looks through all 12 million words and replaces all the vowels with asterisks using regex. To add more load, a “word length” column is added, and then we take words longer than 10 letters and find the most frequent (top 5).

So, in Python this is

df[‘masked’] = df[‘text’].str.replace(r'[aeiou]’, ‘*’, regex=True)

df[‘len’] = df[‘masked’].str.len()

res = df[df[‘len’] > 10][‘masked’].value_counts().head(5)

and this code is executed first through the main processor, then through a GPU.

The main processor (I have the top-tier Intel i9 285k) completes this task in 24 seconds, while the Nvidia RTX 5090 does it in 0.51 seconds. That’s a 46 times difference!

[Pandas CPU] Top Patterns:

masked

s*r w. sc*tt. 23280

s*r t. br*wn*. 23220

j*r. t*yl*r. 16140

bl*ckst*n*. 10860

b***. & fl. 10830

Name: count, dtype: int64

[Pandas CPU] Computation Time: 23.5596 sec.

Transferring data to GPU…

Transfer complete in 1.16s

— Running Benchmark: cuDF GPU —

[cuDF GPU] Top Patterns:

masked

s*r w. sc*tt. 23280

s*r t. br*wn*. 23220

j*r. t*yl*r. 16140

bl*ckst*n*. 10860

b***. & fl. 10830

Name: count, dtype: int64

[cuDF GPU] Computation Time: 0.5108 sec.

TOTAL SPEEDUP: 46.12x

The Maddening Ambiguity of Mathematical Notation | December 02 2025, 15:30

If someone tells you that mathematics is an exact science, don’t believe them. Since I’m currently into data science as a hobby, I’m studying all sorts of things from different books and my brain is exploding at how this can happen in a science where every little detail should fit into a system, otherwise it goes by the wayside. Until it gets to notations. It’s a complete mess there. A set of dialects.

Take, for example, common logarithms. The “standard” for how to denote a logarithm depends on which room of the university you are in. In calculus and number theory, log(x) almost always means the natural logarithm ln(x) with base e. The derivative of e^x equals e^x. It’s “natural”. They’re too lazy to write ln. Yet, where decimal logarithms might appear (like in computer science), log(x) suddenly becomes decimal, and ln(x) is based on e.

The expected value E has an argument in square brackets. Meanwhile, the same square brackets in computer science are used for the step function 0/1.

Or if you see a vector – is it a column or a row? In classical mathematics, a vector is always a column. To multiply it by weights, we write T after the vector and then w for the weights. But in many papers, vectors are thought of as rows. And if you see y = xW+b, then x is not a column, because otherwise the dimensions wouldn’t match up. x here is a row. But in the next paper they write Wx+b. And there x is a column 🙂

Angle brackets . For the dot product, the symbol “⋅” is used, but it is hard to see, especially on a whiteboard, and I very often see that mathematicians use angle brackets for dot product. In general, angle brackets are used for the generalized concept of inner product, where the scalar product is a special case. signifies a certain abstract way to multiply a and b and get a number. Meanwhile, in quantum mechanics this would be written as . And for the scalar product, some use a circle with a dot or x in a circle.

And just for the sake of it, in Russia tangent is tg, while in the USA it’s tan. There’s also tan^-1 and arctan, which are the same, though x^-1 generally means 1/x

Data Science: The Modern Alchemy of the 21st Century | November 16 2025, 04:02

A cryptic post today. While writing a book on RecSys, I caught myself thinking that modern data science is essentially the alchemy of the 21st century. Half of the “best practices” in algorithms lack a solid mathematical framework. It’s a set of heuristics that “just work”. Much like in the 17th century where they mixed everything indiscriminately, it happens now, and if something works better, everyone else starts doing the same. There’s just no answer to the question “why”.

Take, for example, the NCF/NeuMF (Neural Collaborative Filtering) algorithm. The logic goes like this. Say, there are a million movie ratings by users. And 100 million ratings by users yet given – users can’t watch every movie in the world. But out of these 100 million, you need to choose candidates for advertising for a particular user. The algorithm, of course, has a training phase, where weights are calculated, and a prediction stage, where these weights are used on the incoming data.

(What the algorithm does. Essentially, it’s an ensemble of three sub-algorithms, two of which generate their own conclusions, and then their decisions go to a new neural network, the third algorithm, which provides the final recommendation. Smartly, it’s a hybrid of GMF (matrix factorization) and MLP (Multi-Layer Perceptron). The first of these two is based on matrix decomposition, and the second represents a neural network with multiple layers. Weights are adjusted on training data.)

For one positive example, it takes 4 negative ones. Why four? Just because it’s “not too many and not too few”. Would 8 be better? Unknown, but it would definitely take longer to learn.

Why are embedding dimensions 32? or 64? There’s no formula. It’s the “golden mean” between a “dumb” model (few k) and an “overtrained” (many k).

Now about the neural network. Why is the MLP block built as a “tower” (64 -> 32 -> 16)? Why not (50 -> 25 -> 10)? Why ReLU between them (and not tanh for example)? Pure empiricism. The number of layers in the tower is also adjusted.

Why do GMF and MLP parts have different embeddings at the input? Because the authors of the paper tried it, and it “worked out better”. No mathematical proof. Why do they go to the final layer with equal weights? Because they just do.

Why are the outputs of the two paths “concatenated” (concat), and not added or multiplied? “Experience showed that this way the result is more accurate.”

And so it is with everything, up to the choice of optimizer Adam or the “magical” learning_rate=0.001, although at least these have some mathematical basis.

That is, at least a dozen parameters of one algorithm are empirically chosen, with no clear confidence that they are independent of each other. But many of them depend on the dataset, but no one knows how 😉

In general, alchemy.