Jeff will discuss the Brains, Data, Machine Intelligence, Cortical Learning Algorithm he developed and the Numenta Platform for Intelligent Computing (NuPIC).
2. “If you invent a
breakthrough so
computers can learn,
that is worth 10
Microsofts”
3. "If you invent a
breakthrough so
computers that learn
that is worth 10 Micros
1) What principles will we use to build
intelligent machines?
2) What applications will drive adoption
in the near and long term?
Machine Intelligence
4. 1) Flexible (universal learning machine)
2) Robust
3) If we knew how the neocortex worked,
we would be in a race to build them.
Machine intelligence
will be built on the
principles of the
neocortex
5. What the Cortex Does
data stream
retina
cochlea
somatic
The neocortex learns a model
from sensory data
- predictions
- anomalies
- actions
The neocortex learns a sensory-motor model of the world
6. “Hierarchical Temporal Memory” (HTM)
1) Hierarchy of nearly identical regions
- across species
- across modalitiesretina
cochlea
somatic
2) Regions are mostly sequence memory
- for inference
- for motor
3) Feedforward: Temporal stability
Feedback: Unfold sequences
data stream
7. 2.5 mm
Cortex
Layers
Layers & Columns
2/3
4
5
6
2/3
4
5
6
Sequence memory: high-order inference
Sequence memory: sensory-motor inference
Sequence memory: motor generation
Sequence memory: attention
Cortical Learning Algorithm (CLA)
CLA is a cortical model, not another “neural network”
8. 2/3
4
5
6
L4 and L2/3: Feedforward Inference
Copy of motor commands
Sensor/afferent data
Learns sensory-motor transitions
Learns high-order transitions Stable
Predicted
Pass through
changes
Un-predicted
Layer 4 learns sensory-motor transitions.
Layer 3 learns high-order transitions.
These are universal inference steps.
They apply to all sensory modalities.
Next higher region
A-B-C-D
X-B-C-Y
A-B-C- ? D
X-B-C- ? Y
9. 2/3
4
5
6
L5 and L6: Feedback, Behavior and Attention
Learns sensory-motor transitions
Learns high-order transitions
Well understood
tested / commercial
Recalls motor sequences
Attention
90% understood
testing in progress
50% understood
10% understood
FeedforwardFeedback
10. How does a layer of neurons learn sequences?
2/3
4
5
6
Learns high-order transitions
First:
- Sparse Distributed Representations
- Neurons
Cortical Learning Algorithm (CLA)
11. Sparse Distributed Representations (SDRs)
• Many bits (thousands)
• Few 1’s mostly 0’s
• Example: 2,000 bits, 2% active
• Each bit has semantic meaning
Learned
01000000000000000001000000000000000000000000000000000010000…………01000
Dense Representations
• Few bits (8 to 128)
• All combinations of 1’s and 0’s
• Example: 8 bit ASCII
• Bits have no inherent meaning
Arbitrarily assigned by programmer
01101101 = m
12. SDR Properties
1) Similarity:
shared bits = semantic similarity
subsampling is OK
3) Union membership:
Indices
1
2
|
10
Is this SDR
a member?
2) Store and Compare:
store indices of active bits
Indices
1
2
3
4
5
|
40
1)
2)
3)
….
10)
2%
20%Union
13. Model neuronFeedforward
100’s of synapses
“Classic” receptive field
Context
1,000’s of synapses
Depolarize neuron
“Predicted state”
Active Dendrites
Pattern detectors
Each cell can recognize
100’s of unique patterns
Neurons
20. This is a first order sequence memory.
It cannot learn A-B-C-D vs. X-B-C-Y.
Learning Transitions
Multiple predictions can occur at once.
A-B A-C A-D
21. High-order Sequences Require Mini-columns
A-B-C-D vs. X-B-C-Y
A
X B
B
C
C
Y
D
Before training
A
X B’’
B’
C’’
C’
Y’’
D’
After training
Same columns,
but only one cell active per column.
IF 40 active columns, 10 cells per column
THEN 1040 ways to represent the same input in different contexts
22. Cortical Learning Algorithm (CLA)
aka Cellular Layer
Converts input to sparse representations in columns
Learns transitions
Makes predictions and detects anomalies
Applications
1) High-order sequence inference L2/3
2) Sensory-motor inference L4
3) Motor sequence recall L5
Capabilities
- On-line learning
- High capacity
- Simple local learning rules
- Fault tolerant
- No sensitive parameters
Basic building block of neocortex/machine intelligence
23. Anomaly Detection Using CLA
CLA
Encoder
SDR
Prediction error
Time average
Historical comparison
Metric + Time
Anomaly score
CLA
Encoder
SDR
Prediction error
Time average
Historical comparison
Metric + Time
Anomaly score
.
.
.
24. CloudWatch
AWS
Grok for AWS Users
- Automated model creation
via web, CLI, API
- Breakthrough anomaly detection
- Dramatically reduces false
positives/negatives
- Supports auto-scaling and
custom metrics
Customer
Instances
& Services
- Mobile client
- Instant status check
- Mobile OS and email alerts
- Drill down to determine severity
25. Grok for AWS Mobile UI
Sorted by anomaly score
Continuously updated
Continuously learning
26. What Can the CLA/Grok Detect?
Sudden changes Slow changes Subtle changes in regular data
27. What Can the CLA/Grok Detect?
Patterns that humans can’t seeChanges in noisy data
28. CEPT.at - Natural Language Processing Using SDRs and CLA
Document corpus
(e.g. Wikipedia)
128 x 128
100K “Word SDRs”
- =
Apple Fruit Computer
Macintosh
Microsoft
Mac
Linux
Operating system
….
29. Sequences of Word SDRs
Training set
frog eats flies
cow eats grain
elephant eats leaves
goat eats grass
wolf eats rabbit
cat likes ball
elephant likes water
sheep eats grass
cat eats salmon
wolf eats mice
lion eats cow
dog likes sleep
elephant likes water
cat likes ball
coyote eats rodent
coyote eats rabbit
wolf eats squirrel
dog likes sleep
cat likes ball
---- ---- -----
Word 3Word 2Word 1
30. Sequences of Word SDRs
Training set
eats“fox”
?
frog eats flies
cow eats grain
elephant eats leaves
goat eats grass
wolf eats rabbit
cat likes ball
elephant likes water
sheep eats grass
cat eats salmon
wolf eats mice
lion eats cow
dog likes sleep
elephant likes water
cat likes ball
coyote eats rodent
coyote eats rabbit
wolf eats squirrel
dog likes sleep
cat likes ball
---- ---- -----
31. Sequences of Word SDRs
Training set
eats“fox”
rodent
1) Word SDRs created without supervision
2) Semantic generalization
SDR: lexical
CLA: grammatic
3) Commercial applications
Sentiment analysis
Abstraction
Improved text to speech
Dialog, Reporting, etc.
www.Cept.at
frog eats flies
cow eats grain
elephant eats leaves
goat eats grass
wolf eats rabbit
cat likes ball
elephant likes water
sheep eats grass
cat eats salmon
wolf eats mice
lion eats cow
dog likes sleep
elephant likes water
cat likes ball
coyote eats rodent
coyote eats rabbit
wolf eats squirrel
dog likes sleep
cat likes ball
---- ---- -----
32. Cept and Grok use exact same code base
eats“fox”
rodent
33. Source code for:
- Cortical Learning Algorithm
- Encoders
- Support libraries
Single source tree (used by GROK), GPLv3
Active and growing community
Hackathons
Education Resources
www.Numenta.org
NuPIC Open Source Project
34. Research
- Implement and test L4 sensory/motor inference
- Introduce hierarchy (?)
- Publish
NuPIC
- Grow open source community
- Support partners, e.g. IBM, CEPT
Grok
- Create commercial value for CLA
Attract resources
Provide a “target” and market for HW
- Explore new application areas
Goals For 2014
35. 1) The neocortex is as close to a universal
learning machine as we can imagine
2) Machine intelligence will be built on the
principles of the neocortex
3) HTM is an overall theory of neocortex
4) CLA is a building block
5) Near term applications
anomaly detection, prediction, NLP
6) Participate www.numenta.org