datacamp studies

chapter 1: unsupervised learning

• unsupervised learning
  • a class of machine learning techniques used to discover patterns in data
  • opposite of supervised learning which is finding patterns for a prediction task
  • finds patterns in data without a specific prediction task in mind
  • k-means clustering
    • finds clusters of samples (number of clusters should be specified)
    • implemented in scikit-learn
• evaluating a clustering
  • measures the quality of a clustering
  • inertia measures clustering quality - how spread out the clusters are (lower is better)
    • it looks at the distance from each sample to the centroid of its cluster

chapter 2: visualisation with hierarchical clustering and t-SNE

• visualising hierarchies
  • hierarchical clustering is when groups of things form a hierarchy and these clusters are contained in one another
• cluster labels in hierarchical clustering
  • the height on a dendrogram shows the distance between merging clusters
  • a ‘linkage method’ defines distance between clusters
• t-SNE for 2-dimensional maps
  • stands for t-distributed stochastic neighbour embedding
  • maps samples to 2d space or even 3d space

back on the language grind

decided to ditch the anki flashcards for now as they take a while (still useful though)

started my maru hiragana drills at 13.05 and finished them in 4 minutes

the app started throwing in the last section of hiragana characters (yuouon) and it feels god to know that i didn't struggle with identifying them (all correct!!)

started busuu japanese chapter 7 at 14.30 and finished at 15.07

continued on chapter 3 busuu german at 15.10 and finished 15.22

continued on chapter 3 busuu french at 15.27 and finished at 15.45

was too tuckered out to do the frontend coding, maybe tomorrow