jonas-scholz123 Goto Github PK

followers: 11.0 following: 7.0 repos: 29.0 gists: 0.0

Name: Jonas Scholz

Type: User

Company: Imperial College London

Bio: Sofware Engineer with a Background in Theoretical Physics.

Location: London

Hi, I'm Jonas 👋

I'm a German/Jordanian software engineer with a background in theoretical physics from Imperial College London. Currently interested in Rust, meta learning and quantum field theory.

🔧 Tech & Tools

💻 Projects

🚌 London Travel Time Heatmap

This heatmap tells you how long it takes to get to anywhere in London via public transport from one or more starting locations. Ideal for figuring out where to meet with your friends, or how well your potential new home is connected. The backend is written in rust and the frontend in react. Check it out here!.

💥 Particles in a box and ascii animation

Check out my ascii-animated, optimized C++ particle simulation here.

👀 ➡️ 👂 Hearing Images

Another project that I'm proud of is my hearing images application, in which live video data is converted to audio signals, which could function as a translater of visual information for visually impaired people. For a very quick intro, see the gif above. You can check it out (and try it out) here.

🧩 Cryptic Crossword Solver

Over the summer, I contributed to a compsci lecturer's project, a cryptic crossworld solver. It takes a clue from a cryptic crossword, comes up with the most likely solution and explains it to you. For example:

The clue is: County type pair mixed, holding runs (9)
I think the solution is "TIPPERARY"

Let me explain...
I think the answer is supposed to mean "county"
I think "holding" indicates the insertion of one word into another      ◀  helped with this 
   1: The text "runs" resolves to "r"
   2: I think "mixed" indicates an anagram                              ◀  helped with this 
         I think we have the concatenation of 2 sub-clues:
            1: "type" is verbatim text
            2: "pair" is verbatim text
         Joining the results together we get "type pair"
      The required anagram is "tippeary"
For the insertion, "r" needs to be inserted into "tippeary"
The result is "tipperary"

In particular, I worked on a classifier model based on word2vec embeddings, supposed to turn a word into all possible indicators it could represent. In the example above, this classifier evaluates "holding" as an insertion and "mixed" as an anagram indicator. Previously, indicator words were simply referenced against a list of indicators, but using the classifier, even previously unseen words can be classified.

🤝 Get in contact!

Let's make things less one-sided, message me on .

Jonas Scholz's Projects

boomer-humour-exhumer

A PyTorch Image classifier that detects boomer humour.

cart-pole-rl

Using Q-learning on a discretised version of the cart pole problem as a basic reinforcement learning implementation. In the second rendition, I will use a NN to predict Q-values on the continuous state-space (deep reinforcement learning).

complexity

University Project, part of the Complexity and Networks Course 2020

currency-list

:moneybag: List of all currencies with names and ISO 4217 codes in all languages and all data formats.

data-science-textbook

My implementation of the data science methods outlined in "Introduction to Data Science" by Laura Igual and Santi Segui

dialogue-act-classification

do-my-taxes

Does my taxes

emojify

Turn any comment into an emoji-pasta through the power of AI.

forward-forward

An implementation of the paper "The Forward-Forward Algorithm: Some Preliminary Investigations" by Geoffrey Hinton

hearing-images

Allowing visually impaired users to hear images live as they are recorded by transforming greyscale image data into sounds by mapping brightness → amplitude, pixels(x, y) → frequency (using pseudo-Hilbert-curves and Fourier transformations for 2D → 1D mapping)

jonas-scholz123

miscellaneous

Little bits of code that aren't worth their own repo!

mix

Mix datasets easily and extensibly.

msci-project

Master's project: What makes conversations interesting, a NLP approach. We are looking at the structure of conversation under 2 different lenses: as a macroscopic trajectory through a topic space (built from the components of word embeddings) and as a string of microscopic dialogue acts.

msci-thesis-latex

The raw .tex files of my MSci thesis.

networks-project

Implementation of the networks project in 3rd year complexity and networks. This project is part of my university work.

networks_project_rust

This is a repetition of my Network's project (which I wrote in C++) but this time written in Rust. It implements a growing network with uniform probability distribution for new node connections as well as a growing network with a probability kernel proportional to the number of neighbours each vertex has (k).