Semantic Workflow Encoding Using Vector Symbolic Architectures - - PowerPoint PPT Presentation

Semantic Workflow Encoding Using Vector Symbolic Architectures

Chris Simpkin Cardiff University Ian Taylor Cardiff University Graham Bent IBM Research UK Geeth De Mel IBM Research UK Swati Rallapalli IBM Research US

Declarative Approach for Distributed Analytics Self Organising Distributed Analytics ‘Brain Like’ Distributed Analytic Processing for Situation Understanding

Project 5 - Instinctive Analytics in a Coalition Environment

Decentralized Microservice Workflows for Coalition Environments

• Motivation
• Symbolic Vector Representations
• Service Workflows as Symbolic Vectors
• Example of decentralized workflow using symbolic vectors
• Next steps towards the DAIS vision of a “distributed

coalition intelligence” (‘brain’)

Motivation

Objective A common methodology to describe and orchestrate decentralized services to support mission critical data analytics workflow Challenges

• Obtaining a stable endpoint to deploy the service manager is

impractical—if not impossible—due to the variable network connectivity associated with mobile endpoints (e.g., unmanned autonomous systems);

• High latency and cost associated with communication
• Poor infrastructure, especially absent back-end connectivity.

Consequently, in dynamic environments, a new class of workflow methodology is required—i.e., a workflow which

• perates in a decentralized manner.

Intelligent Distributed Analytic Compositions Critical: services that are self-describing and that are discoverable

Images credit: Chris Eliasmith

Hypothesis: Can we use SPA to enable services and data to be self-describing, and compose needed workflows to satisfy requirements?

Semantic Pointer Architecture (SPA)

J Random Permutation Tensor Binding Images credit: Pentti Kanerva John loves Mary

Semantic Vectors for cognitive Services Why Semantic Vectors for Services? Semantic vectors capable of supporting a large range of cognitive tasks:

• Semantic Composition
• Representing meaning and order
• Analogical mapping
• Reasoning

Semantic Vectors are highly resilient to noise (CEMA) and are are good candidates for broadcast communication using HDMAC protocol Semantic vectors have neurologically plausible analogues which may be exploited in future distributed cognitive architectures

Hyperdimentional Binary Symbolic Vectors

Image credit: Chris Eliasmith

Large random binary vectors (n ~ 10000 bits) can be used to create symbolic vectors. Using simple binding operators, sequences of symbol vectors to be represented in a single vector (Semantic Pointer Vector).

Hypercube of n bit semantic vectors with 2n possible vectors (semantic concepts) Distribution of hamming distance for pairs of randomly selected vectors Hamming Distance

s = n-1 Sequence Binding A-> B-> C-> D-> E Pi Binding V = π1A + π2B + π3C + π4E + π5B where π is a random permutation of A XOR Binding V = P0.A + P1.B + P2.C + P3.D + P4.E where P.A = π0P XOR B Sequence Unbinding Pi Unbinding Π-1V = A + π1B + π-2C + π3E + π4B = A + noise XOR Uninding π P0 XOR V = A + noise

Binary Symbolic Vector Capacity

Due to the properties of high dimensional space, vector A will be undecodable when the upper bound of the density of noise vector N approaches the lower. For n = 10,000 and threshold = 10-6 Vector Capacity = 89 symbols

Unbinding of Semantic Vectors

Vector pi unbinding of sequence

• f symbols using circular buffer

Recursive Binding & Unbinding of Semantic Vectors

A B1 B2 B3 B4

C1 C2 C3

D1 D2 D5 Dn

C4

A = T + p00.B11 + p00.p10.B22 + … B1 = T + p00.C11 + p00.p10.C22 + … (p1-1. A1)-1 C1 = T + p00.D21 + p00.p10.D52 + …

• Semantic Pointer Vectors are built from lower semantic levels to higher levels
• Semantic Pointer Vectors are self describing and can be compared with each
• ther
• Semantic Pointer Vectors can then be recursively unbound by vector

unicast/broadcast from highest level

(T) = p0-1.T-1 ; B1 =(p00.B1)-1 (p2-2. A2)-1 C1 = (p00.C1)-1 (p1-1. B1)-1 A1 = (p00.A)-1 = (T) + B10 + p1-1.B21 + … Allows next multiplier to be calculated p1-1p0-2.T-2

Alternative Service Selection Using Semantic Similarity

A B1 B2 B3 B4

C1 C2 C3

D1 D2 D5 Dn

C4

(p00.C1)-1

• Since semantic vectors are self describing and can be compared with each other

alternative semantically similar services are selected when best choice not available.

C2’

A1 = (p00.A)-1 = (T) + B10 + p1-1.B21 + … (p1-1. A1)-1 B1 = T + p00.C11 + p0p11.C22 + … (p00.B1)-1 (p1-1. B1)-1 C1 = T + p00.D21 + p00.p10.D52 + …

Anticipatory Self Provisioning

A B1 B2 B3 B4

• If services can communicate (e.g. multicast/broadcast)

then services at different levels can unbind vectors from above and anticipate when they are going to be invoked.

• Services can monitor the progress of services at level

below and determine progress of the workflow relative to themselves

Look ahead

Hamlet -a Distributed Workflow Example

number of words | 29770 words number of unique words | 4620 words

Interactions at Scene Level

Hamlet as a distributed workflow

Hamlet A1S1 A1S2 A5S2

BS1 FS1 BS2 who’s there ney me answer stand MSn

H = T + p00.A1S11 + p00.p10.A1S22 + … A1S1 = T + p00.BS11 + p0.p10.FS12 + … BS1 = p00.w21 + p00p10.w42 + … w1 w2 (p00.FS1)-1 (p1-1. A1S11 )-1 w4

We have demonstrated how services can learn the play and then recursively unbind in response to the transmission of the Hamlet vector.

(p1-1. H1)-1 H1 = (p00.H)-1 (p00.A1S1)-1 (p00.BS1)-1

Complex Workflows

Pegasus Workflow Generator Montage_20 Workflow DAX Workflow

Complex Workflows

Pegasus Workflow Generator VSA Vector Generation Workflow Vector Generation Recruitment Phase

Complex Workflows

Pegasus Workflow Generator VSA Vector Generation Connect Nodes Phase

Complex Workflows

Workflow Execution

Similarly

Complex Workflows

Workflow Execution

Similarly

S1

Complex Workflows

Workflow Execution

Similarly

S1

Complex Workflows

Workflow Execution

Similarly

S2

Complex Workflows

Workflow Execution

Similarly

S2

Complex Workflows

Workflow Execution

Similarly

S3

Complex Workflows

Workflow Execution

Similarly

S3

Complex Workflows

Workflow Execution

Similarly

S5

Complex Workflows

Workflow Execution

Similarly

Complex Workflows

Workflow Execution

Similarly

Complex Workflows

Workflow Execution

Similarly

S21

Complex Workflows

Workflow Execution

Similarly

S22

Complex Workflows

Workflow Execution

Similarly

Complex Pegasus Workflows

Distributed Workflow in Networks

Evaluation Framework CORE Emulation Decentralized Workflow

Acknowledgement

This research was sponsored by the U.S. Army Research Laboratory and the U.K. Ministry of Defence under Agreement Number W911NF-16-3-0001. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the U.S. Army Research Laboratory, the U.S. Government, the U.K. Ministry of Defence or the U.K.

• Government. The U.S. and U.K. Governments are authorized to

reproduce and distribute reprints for Government purposes notwithstanding any copy-right notation hereon.