lab

This is my central repository for machine learning and AI research and experiments.

Each notebook has links to Colab, Kaggle and Nbviewer; if they can run in Pyodide then they also have a link to my JupyterLite.

Notebooks

• Autograd: "From scratch" iris classifier neural network with Autograd.
• California Housing: EDA and regression with the California Housing dataset.
• EDA: Notes and snippets for exploratory data analysis.
• Functions: Explanations and illustrations of activations, loss functions, and optimization algorithms.
• Monash: Exploring some of Monash's datasets on 🤗.
• Plotly: PX examples.
• Profiling: Demo of Ydata-profiling FKA Pandas Profiling
• Sklearn: A walkthrough of Scikit-learn.
• Sktime: Forecasting sunspots with models from Sktime.
• Statsmodels: Visualizing various datasets included with Statsmodels.
• Sympy: Plotting algebraic functions with SymPy.
• W&B: Experiment tracking and artifact logging on Weights & Biases.
• Widgets: Interactive widget examples with ipywidgets.
• Yellowbrick: Visualizing a Random Forest classifier with Yellowbrick.

TensorFlow

• Autoencoder: Denoising convolutional autoencoder for MNIST.
• CIFAR: Fine-tuning image classifiers on the CIFAR datasets.
• KerasCV: Visualize advanced image augmentations from KerasCV.
• MNIST: Predict handwritten digits including an ipywidgets demo.
• RNN: Time series forecasting and sentiment analysis with RNNs, LSTMs, and GRUs.
• Shakespeare: Multilayer LSTM trained on Shakespeare for character prediction.
• TensorBoard: Watch your tensors flow with TensorBoard and Keras Tuner.

PyTorch

• Iris: Iris classifier with Lighting for training.
• NeuralProphet: Demonstration of forecasting airline passengers with NeuralProphet.

Models

Keras functional implementations from papers.

• AlexNet: ImageNet Classification with Deep Convolutional Neural Networks (Krizhevsky et al., 2012)
• Darknet-19: YOLO9000: Better, Faster, Stronger (Redmon and Farhadi, 2016)
• LeNet-5: Gradient-Based Learning Applied to Document Recognition (LeCun et al., 1998)
• ResNet-18: Deep Residual Learning for Image Recognition (He et al., 2015)
• SqueezeNet: AlexNet-level Accuracy with 50x Fewer Parameters (Iandola et al., 2016)
• VGG-16/19: Very Deep Convolutional Networks for Large-Scale Image Recognition (Simonyan and Zisserman, 2014)

Apps

Gradio

Usage: gradio path/to/app.py

• Iris: Iris classifier and data visualizer with API.
• Todos: Todo app demonstrating session state and custom CSS.

Voila

Usage: voila path/to/notebook.ipynb

• Draw: Drawing app made with ipywidgets and ipycanvas.
• XOR: MLP classifier for the XOR problem.

Installation

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# graphviz
sudo apt install -y graphviz

# ta-lib
wget https://github.com/TA-Lib/ta-lib/releases/download/v0.4.0/ta-lib-0.4.0-src.tar.gz
tar -xvf ta-lib-0.4.0-src.tar.gz
cd ta-lib
./configure --prefix=/usr
make
sudo make install

Usage

To run the JupyterLab server on port 8888:

make

To run the JupyterLite server on port 8000:

make lite

See Makefile for more scripts.

Environment

HF_TOKEN=hf...
WANDB_API_KEY=wb...