Can neural networks become more accurate and efficient by modeling evolution?

Naperville, Illinois
Computer ScienceNeuroscience
$403
Raised of $1,200 Goal
34%
Ended on 11/10/18
Campaign Ended
  • $403
    pledged
  • 34%
    funded
  • Finished
    on 11/10/18

About This Project

Modern AI strategies relying on the Artificial Neural Network (ANN) have achieved massive gains in accuracy and performance over prior methodologies, but they still face flaws in terms of error rates and over-sized, inefficient models. However, natural intelligence has evolved to be accurate and efficient. I hypothesize that if neural network topology design is generated by simulating evolution, then the resulting networks will have higher accuracy and efficiency than existing designs.

Ask the Scientists

Join The Discussion

What is the context of this research?

Modern AI systems perform at high levels on the tasks they were specifically designed to perform, as with the mere 0.21% error on handwritten digit classification achieved at New York University. However, they often fail spectacularly when conditions change even slightly due to the catastrophic interference problem. Part of the problem is that Artificial Neural Networks, the systems behind modern AI, have only a shallow relation to how our brains have actually evolved to function. I hypothesize that AI solutions can broaden their capabilities and make gains in accuracy and efficiency by taking cues from the biological process of group evolution.

What is the significance of this project?

Little data has been collected on how biological mimicry can benefit AI since specific research on the topic only recently began, such as openAI beginning research in march of 2017. As a result, in-depth study is necessary to discern exactly how much performance can be gained, what features can be gained, and by what methods. Specifically, I study the mimicry of group evolution. The proposed experiment will test if group evolution can be used to achieve higher accuracy, lower computational cost, and scalability to any problem size. This data will shed additional light on the validity of biological mimicry, either opening the door to future research or indicating a need to focus on alternative methods.

What are the goals of the project?

My hypothesis is that modeling evolution can increase the efficiency and accuracy of neural networks. I will test the top 20 state of the art designs listed on http://rodrigob.github.io on the standard MNIST data set as well as the naive approaches of the multilayer perceptron and convolutional neural network to establish baseline measurements. From there, I will generate networks with evolution simulations that measure fitness in terms of prediction accuracy and inference time. The algorithm will be tweaked after each trial to improve realism and effectiveness. After 20 trials and iterations of the algorithm, I will compile the data to measure if the evolution simulation improved accuracy and efficiency and note what simulation features were beneficial.

Budget

Please wait...

The RTX 2080 TI serves to vastly accelerate training and inference performance in Tensorflow over the CPU-only version with its 4352 FPUs and new Tensor Core technology. Without GPU acceleration, this research would take several orders of magnitude longer to complete to the point that it lose viability.

Endorsed by

Artificial Intelligence is a key technology for the future. In a time where AI is seemingly everywhere, the technology often fails and progress is stagnating. Seth's novel approaches are something I can believe in and they might well be what we need for the field to progress.

Flag iconProject Timeline

This project aims to have initial findings published by the end of December, 2018 and the results of the completed experiment published by the end of the calendar year 2019.

Sep 26, 2018

Project Launched

Dec 31, 2018

Group evolution initial test results

Mar 31, 2019

Modular component reuse testing complete

Sep 30, 2019

Full system complete and final testing begins

Dec 31, 2019

Results analyzed and published

Meet the Team

Seth Holtquist
Seth Holtquist

Seth Holtquist

I am a computer scientist and AI researcher with a passion for thought experiments. Simply put, I am fascinated by the mind and seek to emulate it so as to better understand how it functions.

I have been interested in computers since my early childhood. I started out playing games and watching videos and became curious about how it all worked as I got older. Over time, I took to watching tutorials and reading guides, eventually working up to disassembling and building several systems. Once I got to high school, my love developed even farther when I realized I could get even more involved with my hobby and solve problems at the same time by getting involved with programming. Since then, I've gone on to learn several languages, create multiple high-end software suites currently in use by my high school's administration, lead our ACSL programming challenges team, and be recognized as computer science senior student of the year. I am now majoring in CS and will make a career out of it.

As for Artificial Intelligence, my interest is much more philosophical. I often run thought experiments or generally think about the natures of consciousness and intelligence and have read many philosophy and psychology books to supplement my understanding. Part of my interest in AI comes from a desire to know more about natural intelligence; to replicate is to understand. Furthermore, I believe that the culmination of human achievement will be our circumvention. It is my opinion that the most impressive, noble, and outright beautiful thing the human race can achieve is the creation of an intelligence greater than our own. Essentially, I view our purpose as to better understand our own thoughts and inner workings in order to spawn the next iteration of intellectual evolution.

Lab Notes

Nothing posted yet.


Project Backers

  • 6Backers
  • 34%Funded
  • $403Total Donations
  • $67.17Average Donation
Please wait...