kato1903 Goto Github PK

I’m Himmet Toprak Kesgin. (That's what I was told.)

I don't like working in the private sector, but I completed the following internships to fulfill academic requirements:

• 2019 - Erlab Teknoloji: (Intern) Training about RESTful Web API

• 2019 - TÜBİTAK BİLGEM: (Intern) Path Planner with Genetic Algorithm and A* algorithm as part of a larger simulation system

• 2019 - YTU CE NLP Lab: (Intern) Research on word embeddings

• 2021 - Today: Research Assistant

  • Yildiz Technical University, Faculty of Electrical and Electronics, Computer Engineering (I'm not like Diogenes yet.)

• 2010 - 2014: Afyon Anatolian High School (76/100)

  • (In our university entrance system, they add your high school grades to your exam scores. With my humble 76/100, I was practically giving everyone else an advance since everyone else is 95+)

• 2015 – 2016: Yildiz Technical University, Department of Foreign Languages

  • (Learned English from scratch this year, proving you can indeed teach an old dog new tricks)

• 2016 – 2020: Yildiz Technical University, Computer Engineering

  • (University entrance degree: 1 out of 93 people. It sounds like a good thing, but if you think about it, you realize I missed the next higher university by 1 person. If you also consider the high school score, meh)
  • GPA: 3.30 / 4.0 (Computer Science Courses A, Hardware Courses C)

• 2020 - 2022: Yildiz Technical University, Computer Engineering (Master’s Degree)

  • GPA: 4.0 / 4.0 (online education 🙂)

• 2022 - Today: Yildiz Technical University, Computer Engineering (PhD)

  • GPA: 4.0 / 4.0 (I was lucky in two final exams where I had a 50-50 chance)

• Languages: Turkish (Native), English (YOKDIL: 93,75) (The local English test only measures reading comprehension and grammar. I still use online translate to read when available.)

• Coding: Python (magic wand) (Pytorch, Tensorflow, Huggingface, any computer automation), Java (good for learning OOP, but I prefer Python), C#, HTML, CSS, Bootstrap, JavaScript, jQuery, Node.js (finished a few Jack Sparrow Udemy courses, able to launch local server and never used it. Good for UI though.), Prolog (I recommend it, it opens horizons), Perl (used for an assignment), MATLAB (used for an assignment), And dozens more (used for an assignment) C (the language I started programming with and recommend), 8086 Assembly (It's hard but should be seen)

• Other: Linux (I use it when there is something I cannot be able to run in Window.), GitHub-Git (I got excited more than ten times but never actually used it.)

We write papers, sometimes pay to publish, and readers even pay to read—can someone explain the logic? Why aren't we all just uploading our works to arXiv to let time and the community sort them out? Truly, why? Moreover, the slow pace of the publishing process and personal biases in peer review often play a larger role in determining the fate of our scholarly endeavors than the actual merit of the work itself. It often feels less like a measure of scholarly merit and more like a game of academic roulette. Why do we play it?

• 2019: Bankruptcy Risk Forecast Based on Company Balance Sheet Data Using Machine Learning (My advisor wanted me to write a paper within the project scope and introduced me to scientific publications, but ordinary conference paper)
• 2022: Development and Comparison of Scoring Functions in Curriculum Learning (Ordinary conference paper)
• 2022: Investigating Semi-Supervised Learning Algorithms in Text Datasets (Ordinary conference paper)
• 2023: Iterative Mask Filling: An Effective Text Augmentation Method Using Masked Language Modeling (It does text augmentation using BERT's mask prediction feature. It's not very good but it has potential.)
• 2023: Cyclical Curriculum Learning (The same dataset and model. Results are better when you use this instead of classical training (model.fit()). It really works and is also my master thesis)
• 2023: Developing and Evaluating Tiny to Medium-Sized Turkish BERT Models (BERT models in Turkish, 5 years delayed)
• 2024: Advancing NLP Models with Strategic Text Augmentation: A Comprehensive Study of Augmentation Methods and Curriculum Strategies (Cyclical Curriculum Learning results are surprisingly good, I have checked several times.)
• 2024: A Robust Optimization Method for Label Noisy Datasets Based on Adaptive Threshold: Adaptive-k (Zero-cost training for label-noisy datasets, no extra model, no 2nd training, no hyperparameters, near-noiseless performance)
• 2024 Introducing cosmosGPT: Monolingual Training for Turkish Language Models (Turkish gpt2 models from scratch 5-6 years delayed.)

• Artificial, Ancient, and Natural Intelligence
• Radical Candor and Honesty (If sincerity is absent so is everything else)
• Natural Language Processing (ChatGPT, LLM)
• Mathematics (The Proofs I Actually Understand)
• Probability and Statistics (Monte Carlo is really good)
• Reinforcement Learning (so fun but hard to code)
• Chess (1800 lichess, 1.e4 best by test, team Nepo)
• Philosophy (Stoicism, Skepticism, somewhere I see a grain of truth)
• Anything to entertain the mind or increase understanding (Irony, Wit, Paradox, Sarcasm, Satire, Humor (Dark), Recontextualisation, implicit understanding)
• As a movie Thriller and mystery, as a series sit-com, as a one-man show stand-up
• Every 2 weeks, spending 16+ hours a day on something (lectures, courses, books, youtube, entertainment).

• The samurai has no goal, only path. (Why ruin it with an ending)

• The more heroes we have, the less we think.

• Not any kind, to best of my efford.

• ask, “Why is this so unbearable? Why can’t I endure it?” You’ll be embarrassed to answer.

• None. Except for anonymous accounts I use to look at memes.

• Man's logic: Claim the moon but still fight over parking spaces. No man's land is every man’s peace.
• Rebels thrive on the rules they refuse to obey.
• Saint is just a sinner who keeps trying.
• Solve with intellect < prevent with foresight.
  Ease with tech < make it irrelevant.
  Open doors with money < exist beyond doors
  Lead well < let self-govern.
  Create order < transcend rules.
• Every agreement actually proves that two people do not think the same thing.
• Self-censorship is carrying out someone else's dirty work by our own will.
• Men dig tons of earth to find an ounce of gold.
• The true challenge in computing is not making machines think like humans but making humans think about the potential of machines.
• Shadows are not merely absence of light, but companions proving its presence.
• A mirror may be spotless, yet it holds no power to cleanse the visages it mirrors.
• Palaces boast myriad windows, yet offer little in the way of vision.
• Social rules are like a spider's web; even the strongest one will break one day.
• The more you observe, the more you realize how little you observe.
• If you know the end, it's a chore.
• Men build walls, not to protect treasures, but to hide their emptiness.
• The mightiest empires fall, but a single act of kindness echoes through eternity.
• While kings sit on thrones, ideas rule the world.
• In the market of desires, I'm the one asking for refunds.
• The mightiest king still fears the honesty of a child.
• Society is strange: They ask for your opinion only to get offended by it.
• Fashion dictates our beliefs more than reason, and we choose assumption over understanding.
• The world doesn't need more millionaires; it needs more Diogenes.
• Companies exploit, society chains, people conform; change whispers.
• You’ll find being honest and transparent puts you at risk—of becoming well-respected, that is.
• If you always cater to consensus, you’re following, not leading.
• Groups search for consensus, individuals search for truth.
• Man's true voyage doesn't lie in seeking new landscapes, but in having new eyes.
• Doubt is the surest companion.
• Stubbornness is a trait that can only be validated by another stubborn person.
• Lost shoes, found path
  Lost map, found world Lost Wallet, found worth Lost sight, gained vision Lost name, found self
• If you overvalue possessions, people begin to steal.
• Even if everyone is looking; do the right thing.
• Justification itself is another belief.
• The worst is he who treats his slave well; he has simply found a better whip and perpetuates the system.
• Steal enough and manipulate belief, you could buy a communitys favor to canonize you. But beware for you’ve just funded the next revolution.
• Without justice, a state is just a larger gang; with it, it becomes an ideal of self-governance.
• The modern economy creates the illusion that wealth feeds poverty, when in fact poverty feeds wealth.
• Volcanoes are silent until they burn the earth itself.
• I invite you to evade the rational herd that, in their escape from the flock, forgot they too have become one.
• Chasing endless wants, facing growing discontent: Want is the treacherous mirage, promising life but bringing decay.
• Who would applaud a man perfectly acclimated to a society that is but a hospice of moral illness?
• Midst lies, insincerity, misread intentions brewing resentment, dealing obscured coercion yet within us lies nobler potential, dear mortals.

Diogenes when asked letters of recommendation:

When he sees the kind of man you are, he will understand. If he has the ability to distinguish between good and evil, he will know whether you are good or bad. However, if he doesn't already understand these concepts, writing a thousand letters to him will still not make him understand.

• Email: tkesgin at yildiz.edu.tr (if you pass the spam filter of the email provider and my brain’s spam filter, I will reply)