Ideas

The portfolio of ideas

Each idea is a research line with a question, what has been tried, and how far it has got. Where an idea has something you can open, its lab is linked. No idea is ranked above another.

1. BuiltThe root

   Irregular series as trees of shards

   Question Can a series be read by its shape, without forcing it onto a grid?

   Every sample is kept where it fell. The series is cut at its turning points into shards, and the shards nest into a tree at every scale, with graph iterated function systems (GIFS) as the mathematical footing. Available as a built, metered API, and offered as an alternative to candlesticks.

   Shards, not candles →3 write-ups
2. LabShapes in series

   Search by shape, at any scale

   Question Can a shape in one series be found in any other, whatever its size?

   Every subtree of every series stored as a graph with exact signatures and subtree vectors, so a shape drawn round part of one series can be found again in any other, at any price and any time scale.
3. ExperimentShapes in images

   From 1D series to 2D images

   Question What does a tree of shapes look like over two dimensions?

   The same tree over two dimensions of position: two constructions compared, quadtrees with sloped and shared-corner tiles, tested on real photographs, with a shift test.
4. ExperimentShapes in series

   Shape as a format

   Question Can a whole history travel as a small file and be drawn at any detail?

   Histories served as small GIFS files from a CDN, drawn at whatever detail a widget has room for: a sparkline, a hover card, a full chart, down to single days.
5. ConceptShapes in series

   Reading new data by precedent

   Question Can today’s shape, matched against the past, forecast, flag or explain?

   Foresight, Sentinel and Essence: forecasting, anomalies and patterns read by matching today’s shape against the past. Concept sketches on synthetic data.

   Concept figures →
6. ExperimentShapes in series

   Faults in sensor history

   Question Can slow faults be seen in unevenly recorded sensor data before an alarm?

   Years of unevenly recorded sensor readings read by shape, to see slow faults before an alarm limit is reached. A benchmark on public data, written up.

   2 write-ups
7. ExperimentShapes in images

   Shape trees as compression

   Question How does an entropy-coded shape tree compare with JPEG and JPEG 2000?

   An entropy-coded tree of shared-corner tiles, measured against JPEG and JPEG 2000 on the same images.
8. LabShapes in knowledge

   Knowledge an agent can cite

   Question Can every fact an agent gives be traced to the passage it came from?

   Documents read into a graph typed by an ontology, every fact traced to its passage, and an agent that cites them. Live on S&P 500 earnings, with the architecture open source as Knowledge Store.

   Earnings Lab →3 write-ups
9. ExperimentShapes in knowledge

   Ontologies for whole industries

   Question Can an industry standard be distilled into an ontology that drives extraction and chat?

   Industry standards distilled into ontologies that drive extraction and chat: banking service domains, a retail banking subset of FIBO, and a schema for frontline knowledge.
10. ExperimentKnowledge and markets

    Earnings read-through

    Question Does one company’s earnings release predict how a linked company’s shares move when it reports later in the season?

    Economic links taken from the knowledge graph, tested across reporting seasons.

    Supplier links in the Earnings Lab →