← All workshopsRead Paragraph Vectors

Natural Language Processing

Give a document its own memory

BOTIE

Multiply the positive-word prediction error by the frozen word-prediction vector.
Ready to test
0 passed · 0 failed · 1 not run

An empty floor. Start with the information your machine needs.

Received result

Your answer will arrive here.

Examples, experiments and workspace controls
Live values

Connect a tool's inputs to see what it produces. You can try a branch before connecting it to Result.

Try your own inputs

Use a copy of the example to try your own data. Keep the problem's assumptions in mind, and predict the effect of a change before checking the output. You can also add another branch to compare two ways of doing the calculation.

Start at an output arrow, then click or drag to an input. Drag empty space to pan. Delete removes a selection. Ctrl/Cmd+Z undoes a change.
Context for this challenge

Multiply the positive-word prediction error by the frozen word-prediction vector.

A new document needs a representation in a space the model has already learned. Let's adjust its document vector using one observed word while keeping the word-prediction weights fixed. Repeating that process would infer a vector for the document without retraining the whole model.

Inspect the component calculation
Multiply(
  Value: Subtract(
    Starting value: Sigmoid(
      Score: Dot product(
        First vector: Current document vector
        Second vector: Frozen word-prediction vector
      )
    )
    Value to remove: 1
  )
  Multiplier: Frozen word-prediction vector
)
Examples and results
Current document vector
  1. 00
  2. 10
Frozen word-prediction vector
  1. 01
  2. 12
Required result
  1. 0-0.5
  2. 1-1
Try a fresh construction

This clears this draft, including its undo history. Earlier successful attempts stay recorded. Build again without opening hints to record an independent attempt.

Stuck? Inspect a worked construction

The tools below form one possible solution. You can assemble it, follow its signals, and undo that change to return to your own construction.

Assumptions and test precision

This is a small construction problem. The lesson explains the assumptions and the parts of the full method that this example leaves out. Numerical tests allow a difference of 0.001. Lists and structures must also have the right shape.

Your last tested pieces can start the full construction. An existing draft there keeps its own pieces.

Infer a document from a neutral prediction

This check uses the pieces from “Update the document rather than the word weights”.

BOTIE

A new document's two coordinates cancel in the observed word's score. Take one positive-target inference step with learning rate 0.2. Leave the word-prediction weights frozen.
Current document vector
  1. 01
  2. 1-1
Frozen word-prediction vector
  1. 01
  2. 11
Inference learning rate
0.2

Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.

Build a construction in the workspace above first. This case will test those pieces.

Which input stays frozen, and which is being inferred for the new document?

Inference can involve optimisation when a representation must be fitted to new observations.

This is a small construction problem. The lesson explains the assumptions and the parts of the full method that this example leaves out.

Work through the Python challenges →
Continue the courseTransformer Blocks →
← Pooling a Text