Indie's Substack

Indie's Substack

Home
Notes
Archive
About

Part 3: Model Architecture as Context Aggregation

Why BiLSTM with Attention Works — and Where It Breaks
READ THE LATEST
Part 2: Training Data Design for Missing Character Prediction
How Synthetic Corruption Quietly Defines Performance Limits
 • Indie Quant
Part 1: Framing Missing Character Prediction Correctly
Why Character Prediction Fails Before Training Even Begins
 • Indie Quant
Part 3: Text Cleaning, Model Scaling, and Knowing When to Stop
How far sentiment analysis can be pushed — and why it eventually plateaus
 • Indie Quant
Part 2: Feature Geometry, Representation Limits, and Baseline Models
Why better representations outperform better algorithms in sentiment analysis
 • Indie Quant
Part 1: Framing Sentiment Analysis as a Systems Problem (Not a Modeling Problem)
Why most sentiment analysis pipelines fail before training even begins
 • Indie Quant
Part 3: Using Silhouette Score to Choose the Right Number of Clusters in Time Series Clustering
How silhouette score reveals the optimal number of clusters in shape-based time series clustering
 • Indie Quant
Part 2 - Correlation vs Euclidean Distance: Seeing Time Series Similarity the Right Way
Why Euclidean distance fails and correlation distance reveals true structure in shape-based time series clustering
 • Indie Quant
Time Series Clustering: Why Distance Metrics Fail Before You Even Cluster
How similarity measures like Euclidean and Correlation distance can help us in solving real-world time series clustering problems
 • Indie Quant
SARIMA Forecasting in Practice: auto-ARIMA, Ensembles, and Production-Ready Demand Forecasting (Part 3)
How to tune SARIMA models, convert daily forecasts to hourly demand, and use ensembles to squeeze out RMSE gains
 • Indie Quant
Indie's Substack
Indie's Substack
My personal Substack

Indie's Substack

AboutArchive
© 2026 Indie Quant · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture