Various machine learning case studies, created as educational materials or showcases.
- MAPE regression (training in Keras + geometric insights)
- Mixture of Experts (fitting a mixture of models, such as NNs...)
- Malfunction Likelihood by Recurrent Neural Networks, a case study on device failure temporal data
- Anomaly Detection by Recurrent Neural Networks in Time Series Families (intraday data)
- Random Embeddings vs PCA as Feature Extractors
- Piecewise Linear Models trained by Bayesian Inference
- Robust Feature Selection on Heart Disease Dataset
- Autoencoders help in small dimension, case study on Framingham Heart Study Data
- Hessian approximation for Neural Networks
- Employee Attrition Model trained in Tensorflow and productionalized on Google Colab with REST API
- Heat Waves Models, trained with various Python frameworks (statsmodels, GPFlow, PyMC3) and used for complex hypotheses testing
- Hiring Bias Study, demonstrates the superiority of model-driven statistical insights over exploratory findings.
- Text Clustering on GPU, implements and optimizes a modern text clustering pipeline (embeddings, dimension reduction, clustering) on GPU.
- Reinforcement Fine-Tuning of LLMs for Emotion Recognition implements fine-tunning of OpenAI models for better emotion recognition.
Some useful recipes