About Me

عربي

I am Kamal Al-Ameri, an Iraqi 16 year old "aspiring/independent researcher". My approach to learning is very hands-on, as I prefer to tackle problems that seem far ahead of me to expand my horizons. I am passionate about science in general, but I am particularly focused on quantum computing, spiking neural networks, and high-performance computing. However, I often pick up projects from outside those fields, partially out of passion, and partially because insights from one field often become helpful in a seemingly unrelated field.

A Selection of My Work

  1. RQISM: The name stands for Rust Quantum Information SiMulator, and it is a rust-based multi-backend quantum circuit simulator with a primary focus on performance. The statevector simulator has achieved simulation of the preparation of the 20-qubit GHZ state in only 44 microseconds, which in comparison to Google's simulator, cirq, is about 720 times faster than cirq.Simulator().simulate(), on the same hardware and operating system. This was intially meant to be a small, 2 week project to learn quantum computing on a deeper level, but spiraled over the course of a year into highly optimized simulator written just to satisfy curiosity.

  2. thimni: A fast, n-dimensional, linear algebra library agnostic implementation of my novel algorithm for collision detection between signed distance functions. An informal PDF explaining the algorithm is provided. I have written a demo demonstrating a small 3d first-person game where the player is a capsule and the scene is a large, pulsing fractal, that the play may destruct in real-time. I made this because I find SDFs to be very useful and powerful, but found that their utility was limited by the non-existence for SDF collision algorithms that were both efficient and generalized.

  3. SQ-SVM: A highly critical analysis of a 2019 quantum machine learning paper, in which I mathematically prove that the circuit proposed in the paper for kernel generation is entirely redundant, and that the same result can be achieved with 1 qubit. I then demonstrate this on the MNIST dataset and the boston housing dataset. An informal PDF containing the proof and demonstration is available. The project was driven by a simple desire to fix an apparent problem.

  4. LSNN: The newest project listed, a cache friendly spiking neural network that mimics the locality of the overwhelming majority of real neurons, and exploits that locality to ensure extremely fast iteration during the two processes powering SNNs: spike propagation and STDP, by creating specialized data structures for storing data about synapses. The improvement was measured by a new specialized metric I called "anti-contiguity". This was an example of insight from one field helping in another, as my hobby of game development as a creative outlet is where I gained my knowledge about many of the concepts utilized in this project.

Failures

I have succeded with many of the problems I have tackled, but failed with some, including (because I want to return to them in the future):

  1. Quantum circuit synthesis from irreversable classical functions using genetic algorithms (results)

  2. Simulation of Dr. Erik Verlinde's entropic gravity model, and generally simulation of AdS/CFT correlation.

Contacts

Email: kmalalamry8@protonmail.com Discord: laxative2009