Data

Tools

Indexes

Applications

Experiments

Open Source Science

mei-reconstruction

Code for my BCs thesis on mice's most exciting images

https://github.com/fcantatore/mei-reconstruction

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (6.3%) to scientific vocabulary

Keywords

cnn-for-visual-recognition digital-twin most-exciting-input
Last synced: 6 months ago · JSON representation ·

Repository

Code for my BCs thesis on mice's most exciting images

Basic Info
  • Host: GitHub
  • Owner: fcantatore
  • License: mit
  • Language: Jupyter Notebook
  • Default Branch: main
  • Homepage:
  • Size: 19.2 MB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Topics
cnn-for-visual-recognition digital-twin most-exciting-input
Created over 1 year ago · Last pushed over 1 year ago
Metadata Files
Readme License Citation

README.md

MEI-Reconstruction

This repository contains the code used in my dissertation Most Exciting Images in a CNN reconstruction of the mouse brain requested by the BCs of Mathematical and Computing Sciences for Artificial Intelligence (BAI) at Bocconi University, Milan.

Abstract

This thesis explores the optimization of images as best stimuli for neurons in the visual cortex of a mouse. I will present the structure of the optimization algorithm used to develop these Most Exciting Images (MEI) and implement it in a control environment to verify its behavior before applying it to the main model.
The main model used to simulate and predict neuron activation is a Digital Twin with a CNN as its core architecture and trained on the comprehensive MICrONS dataset.
I will explain what challenges I encountered in both the control environment and the main model, what solutions I adopted and compare their effectiveness. Finally I'll expose my findings along with a sample of the MEIs developed and my considerations of the current way to model neuron activation for simple cells in V1 based on them.

Structure

  • Control Environment; contains code relative to challenges, solutions and results encountered in the control environment and used as an initial validation of the results.
  • Main Model; contains the code used to obtain the MEI as well as animations of the optimization process and both penalties on Contrast and Brightness.
  • animations; is a folder containing examples of the animaited optimization process, from the random noise to the final MEI.
  • images; is a folder containing both individual MEIs and collections of the most relevant in terms of oracle score.

Examples

Owner

  • Login: fcantatore
  • Kind: user

Citation (CITATION.cff) 

# This CITATION.cff file was generated with cffinit.
# Visit https://bit.ly/cffinit to generate yours today!

cff-version: 1.2.0
title: MEI-Reconstruction
message: >-
  If you use this software, please cite it using the
  metadata from this file.
type: software
authors:
  - given-names: Fabio
    family-names: Cantatore
    email: fcantatore02@gmail.com
identifiers:
  - type: url
    value: >-
      https://github.com/fcantatore/MEI-Reconstruction
    description: >-
      Python notebook with implementation of numerical
      methods
repository-code: >-
  https://github.com/fcantatore/MEI-Reconstruction
abstract: >-
  Updated code implementation for the reconstruction of Most Exciting Images
  of a mouse brain using a CNN architecture, part of Fabio Cantatore's
  bachelor thesis at Bocconi University in the degree of Mathematical and
  Computing Sciences for Artificial Intelligence.
date-released: 2024-06-13

GitHub Events

Total
Last Year