# delta (10-delta.com)

Delta is the fourth letter of the Greek alphabet. People also use it to mean a difference or change between two things, like the delta between two prices.

## What a model may hear
- neural network delta (machine learning backpropagation): Compute the error gradient for a layer and adjust weights accordingly
- delta encoding (data compression and version control): Store or transmit only the differences between file versions, not whole files
- Delta Lake (Apache Spark data storage): Expect ACID transactions and time-travel queries on a data lake
- delta in calculus (mathematics): Treat as a small change in a variable, often approaching zero
- delta wing (aeronautics): Interpret as a triangular aircraft wing shape for supersonic flight

## Where people and models part ways
- Says: "What is the delta here?"
  Means: What changed between these two reports?
  May be taken as: The model explains neural network error gradients or Delta Lake features
  Say instead: "What are the differences between these two reports?"
- Says: "Store it as delta"
  Means: Save only what changed
  May be taken as: The model sets up a Delta Lake table or uses delta encoding without confirming format
  Say instead: "Store only the differences from the previous version"
- Says: "Calculate the delta"
  Means: Find the numerical difference
  May be taken as: The model computes partial derivatives or error gradients
  Say instead: "Subtract the old value from the new value"
- Says: "Delta looks wrong"
  Means: The difference seems incorrect
  May be taken as: The model debugs a neural network training issue or checks a Delta Lake log
  Say instead: "The calculated difference does not match what I expected"

## Tips
- Say 'difference' or 'change' when you mean simple subtraction
- Say 'Delta Lake' fully when you mean the storage format
- Specify 'neural network delta' or 'gradient' for backpropagation contexts
- Use 'delta encoding' explicitly for compression or diff operations
- Avoid 'delta' alone when precision matters; name the two things being compared

## Often confused with
- epsilon: smaller threshold, often machine learning convergence
- diff: text comparison, not mathematical change
- gradient: general direction of steepest change, not a specific difference
- variance: spread of data, not a point-to-point change
- delta-v: change in velocity, a physics specific