Recent Stock Market Pictures & Work in Other Domains
Summary Points:
1) Actionable Domain Insights Generated
2) By Changes in Architecture, Software, Data, Metrics
A Typical View of Stock Value at 30 Second Blocks - Linear Metric - Averaging
Same Collection of Stocks at Finer Level Detail
Using Non-Linear Similarity Profile Metrics
Finer Level Details in Smaller Time Range Above;
Below is larger Time Span with Same Stocks
Domain Level Insight Into Market Movement
Creating Actionable Opportunity
Below A Very Small Timeline Segment of Stock trades within a 1 second band
Common Practices Average 1 second values - erasing true Insight
Other Domains Also Average Data Sets
Below are Pharma Genomic Interaction Correlations
Non Linear Metrics in Upper Triangle - Linear in Lower Right Triangles
Deeper Implication is that world is missing opportunities
Other Metrics Exist Providing Better Insight and Efficiency
----------------------------------------------------------------------------------------------------
Larger Industrial Domains Insight - Using Alternative Metrics Sets
$100 M / mo - 10 x Return On Investment
Bought Out by Intel ( / Chip ) Manufacturer
Above Business Requirements, Solution that could answer any commercial aviation domain question
Capability: Large Compressed Graph Index of all Domain data sets
Associated in triple space by Entity - Attribute - Entity Chaining
A set of non-linear similarity metrics that could determine similar profiles over large graphs
Questions Answered included:
Customer Ops Center - Airplane On Ground Problem Reports
What parts a specific customer is interested in most - cost, time ?
What Designs ? What Structures ? What Plane Models ? From What Line ?
What Customers had similar operational requirements ?
What set of parts do I need to have in inventory now, next week ?
How Many and where needed ?
What Tools are needed ?
What engineering staff do you need to fix ?
The Inputs were:
All - Parts, Designs, Drawings, Customers, Airlines, Line numbers, Aircraft Numbers,
Part Numbers Mapping to Function
(ATA-2 - System, ATA-4 - Subsystem, ATA-6 -Assemblies, ATA-8 - LRU)
PtolemyII Modeling and Simulation Tool - Linked System Blocks
Varying Inputs - Assess Parameter Sensitivity in Signals Vs Noise Studies
-----------------------------------------------------------------------------------------------
In older tradition, experimental evidence provided basis to determine model parameters
Below is Alternate Data Driven Deep Learning Predictive Model
Below Detailed View of Non-Linear System Degradation within a Jet Engine Turbine

Figure Generated by Patrick Harris using GPU, as an alternate to Dr. Wagner's infrastructure; Dr. Wagner's infrastructure scales better and less expensive than multiple GPU's. More Details here - if you were to ask me during a review.
Figure Generated by Dr Alan Wagner in consultation with Dr Paul Hoffman and Patrick Harris
Comments
Post a Comment