How sandromikautadze/neural-response-identification works
Project type and purpose
• This is a data‐analysis and modeling project (not a web or full-stack app). Its goal is to build and compare mathematical models (linear, quadratic, MLP) that predict mouse VISam neuron spike counts from static grating features.
File structure
• data/
• CSV files (static_gratings_VISam_*.csv) – the raw/cleaned datasets of stimuli and spike counts.
• utils/
• regression.py – utility functions for building and evaluating regression models.
• Jupyter notebooks at root
• data_analysis.ipynb – exploratory analysis and visualization.
• neurons_range_selection.ipynb / neurons_variance_selection.ipynb – separate modeling pipelines based on different neuron selection criteria.
• report.pdf – compiled write-up of methods, results, and conclusions.
• Configuration files
• .gitignore – standard ignore rules.
README insights
• Research question: predict spike counts given orientation, spatial frequency, phase.
• Workflow: select responsive neurons → featurize inputs (linear, quadratic, sinusoidal) → fit models (MLP, linear) → evaluate → report.
• No external services; everything is local files and Python notebooks.
Main components
a. Data Layer
– Raw/cleaned CSV files in data/
b. Preprocessing & Feature Engineering
– Python functions in utils/regression.py
– Notebooks call these to build feature matrices (linear, quadratic, sinusoidal)
c. Modeling
– MLP and linear regression routines in utils and notebooks
d. Evaluation & Selection
– Range-based and variance-based neuron filters (notebooks)
– Performance metrics calculation
e. Reporting
– Figures and tables embedded in notebooks → final report.pdf
Relationships and data flow
– The notebooks orchestrate calls to utils/regression.py.
Architectural patterns & design principles
• Modular separation of concerns:
• Data storage (data/)
• Reusable code (utils/)
• Notebook workflows (analysis & modeling)
• Documentation output (report.pdf)
• Reproducibility via notebooks and version control.
• Labels: describe what each component does (e.g., “compute quadratic features”).