Data & ML

All posts in Data & ML. Korean originals are linked until English translations are approved.

Machine learning, data work, and applied AI foundations.

• Data & ML HalluLens Paper Review

  Forget hallucination benchmarks plagued by concerns about existing data leakage!

• Data & ML batch vs batch size

  A bundle of data samples—the number of samples contained in one batch

• Data & ML The Concepts of Embeddings and Tokenization

  Dividing a sequence (such as text or f0 audio divided into frames) into tokens, quantifying them, and mapping them through indexing; arranging those tokens based on their semantic associations

• Data & ML Tensor Manipulation Tips - What is unsqueeze?

  unsqueeze is a function in PyTorch that adds a new dimension (an axis of length 1) to a tensor. For example, if the shape of quantized_f0 is (1703,), quantized_f0.unsq

• Data & ML What is nn.Embedding

  Motivation\nMy understanding of the concept of embedding is that natural language or entities are numerically placed in an embedding space according to the semantic similarity desired by the user.\n\nBut rather than passing natural-language lyrics directly through nn.Embedding, are the lyrics already v

• Data & ML Why do we transpose when using nn.Embedding?

  The shape after summing the lyrics encoder and melodyU encoder is (1, 192, 1) - (b, h, time(seq_len)). These undergo FFT operations again in the enhanced condition encoder, but likewise nn

• Data & ML Mean pooling - Let's match the embedding dimensions

  \nProblem\nFFT encoding Lyrics and the quantized f0 melody produces [batch size, hidden channel, time(seq_len) -> sequence length].\nThe sequence lengths might be different, right??\n\nSolution\nm

• Data & ML Overfitting with k-fold cross-validation

  Hyperparameter tuning and CV overfitting

• Data & ML How to Identify Outliers

  https://esj205.oopy.io/72782730-23e4-43cf-8799-f3cdcbcb57b9

• Data & ML What Is a Quartile? (What Is a Box Plot?)

  Whisker plot

• Data & ML Multilabel Classification Fine-Tuning

  Experimental design evaluation metric macro

• Data & ML RAGchain Benchmark: 3-Line Summary

  Brief benchmark summary

• Data & ML What is context length?

  It is the number of tokens an LLM can process at once. It is the maximum length of an input sequence for an LLM. A nice explanation. This is important because 1. A model cannot exceed the context length when summarizing articles. 2. Long-term planning ta

• Data & ML What Are Qrels?

  Do you know Dr. Query?

• Data & ML RAGchain benchmark Mr.tydi

  Do you know Mr. TyDi?

• Data & ML MRR metric

  What is MRR? (mean reciprocal rank)

• Data & ML RAGchain Glossary

  I’m organizing this place for my own reference because I keep coming across a barrage of unfamiliar words while reading papers.

• Data & ML Blending

  Delicious coffee blending—apparently the No. 1 algorithm in the Netflix prediction competition??

• Data & ML Hyperparameter

  https://rosypark.tistory.com/236

• Data & ML PCA (Principal Component Analysis)

  What is PCA?

• Data & ML Overfitting & Underfitting

  What is overfitting?

• Data & ML Supervised Learning (Classification)

  Let's learn about classification

• Data & ML Supervised Learning (Regression)

  Let's learn about regression.

• Data & ML Data preprocessing

  Data preprocessing process

• Data & ML Forms of Data

  Participation in the Military AI/SW Competency Enhancement Program