How abhay-bharat/neural-network-mi-assignment-3 works
Project type and purpose (from file tree + README)
This is a machine learning / educational assignment project: “Implementation of basic neural network using numpy”. It is not a deployed service; it’s a local, notebook-driven workflow with:
• A raw dataset (LBW_Dataset.csv)
• A cleaned dataset (cleaned_LBW_Dataset.csv)
• Jupyter notebooks that perform data cleaning and model experimentation (cleaning_data.ipynb, nn.ipynb, plus temp notebooks)
• A core Python implementation of the neural network (Neural_network.py)
• Supporting artifacts (Assignment_3_-_Machine_Intelligence.pdf, debug.log, .ipynb_checkpoints)
Data layer (files)
• Raw dataset: LBW_Dataset.csv
• Cleaned dataset: cleaned_LBW_Dataset.csv
Preprocessing / data preparation
• Notebook: cleaning_data.ipynb
• Reads raw CSV
• Performs cleaning/transforms/feature prep
• Writes cleaned CSV (inferred from presence of cleaned file)
(Also present: temp_file.ipynb and checkpoint versions; these are non-architectural artifacts unless they contain essential steps.)
Model / core logic
• Neural network implementation: Neural_network.py
• This is the key “library/module” in the project.
• Likely contains forward pass, backprop, weight updates, loss, etc. (typical for numpy NN assignments).
Experimentation / training / evaluation driver
• Notebook: nn.ipynb
• Loads cleaned data
• Uses Neural_network.py to train/evaluate
• Produces metrics/results/plots (even if not committed as files)
Documentation / assignment spec
• README.md (minimal)
• Assignment_3_-_Machine_Intelligence.pdf (the spec that dictates requirements)
Tooling/auxiliary artifacts (usually not in the ma
• .ipynb_checkpoints/* (Jupyter internals)
• debug.log (runtime artifact)
Inputs
• Raw dataset CSV
• Assignment PDF (requirements)
Preprocessing stage
• cleaning notebook (and key transformations, if known)
Training/Evaluation stage
• nn notebook (driver/orchestrator)
Core Model Implementation
• Neural_network.py (as a reusable module)
Outputs
• Cleaned CSV
• Evaluation metrics/plots (conceptual)
• (Optional) saved weights/models if the notebook saves them (not shown in tree, so mark as “optional / not present in repo”)
What technologies/frameworks to annotate
Even if the README is minimal, the repo strongly suggests:
• Python
• NumPy (explicit in README)
• Jupyter Notebook (the .ipynb workflow)
• “NumPy (confirmed)”
• “pandas/matplotlib (if used in notebooks/code)”
“Don’t overthink it”: split into as many components as possible (without inventing)
• Data normalization/encoding utilities (if present)
• Layer definitions (weights/bias)
• Forward propagation
• Loss function
• Backprop/gradient computation
• Optimizer/update rule (e.g., SGD)
• Prediction/inference
• Metrics (accuracy, confusion matrix, etc.)
• Train/validation split
• Hyperparameter selection
• Training loop orchestration
• Evaluation/visualization
• Missing value handling
• Feature selection/encoding
• Scaling/normalization
• Output schema definition (columns of cleaned CSV)