The document provides an overview of digital image processing analysis in industry and machine learning. It begins with an introduction of the speaker and their background. It then defines machine learning as algorithms that improve performance on tasks through experience. The document explains concepts like training models on data, scoring new data, deep learning using multiple neuron layers, and common machine learning algorithms. Finally, it dives deeper into convolutional neural networks (CNNs), how they work using examples, and the mathematical operations behind feature matching in CNNs.
3. “A computer program is
said to learn from
experience E
with respect to some
class of tasks T
and performance
measure P,
if its performance at tas
ks in T,
as measured by P,
improves with
experience E”
4. 12/13/16Intel Confidential 4
What is Machine Learning
4
Machine Learning, a key tool forAI, is the development, and application of algorithms that improvetheir
performance at some task based on experience (previousiterations)
Training: Build a mathematical model based on a data set
Scoring: Use trained model to make predictions about new data
Deep Learning
Algorithms where multiple layers of neurons learn
successively complex representations
RBM …RNNCNN
Statistical/ OtherMachineLearning
Algorithms based on statistical or other techniques for
estimating functions from examples
GA
Linear
Regression
SVM
NaĂŻve
Bayes
5. 12/13/16Intel Confidential 5
End to End Workflow
5
Things
Data Annotation
Label & Prep Data
Model Deployment
Over-the-Air
Secure & Real Time
Model Scoring
App Dev & Runtimes
Embedded OS
Model Update
Track Model Drift
Manage Model Lifecycle
New Data
New Model
Data Aggregation
Data Curation
Catalog Data Sets
Model Training
Train forAccuracy
Model Validation
Run Simulations
Cross-Validate
Distributed SystemsData Acquisition
Model Performance
Tune for Performance
Analyze Longitudinal Effects
6. 12/13/16Intel Confidential 6
What is DeepLearning
6
Multiple definitions, however, these definitions have in
common:
• Multiple layers of processing units
• Supervised or unsupervised learning of feature representations in
each layer, with the layers forming a hierarchy from low level to high
level features.
7. 12/13/16Intel Confidential 7
What is CNN
7
Essentially neural networks that use convolution in place of general
matrix multiplication in at least one of their layers
9. 12/13/16Intel Confidential 9
How CNN Works
9
A toy ConvNet: X’s and O’s
X or OCNN
Says whether a picture is of an X or an O
A two-dimensional
array of pixels
14. 12/13/16Intel Confidential 14
How CNN Works
14
-1 -1 -1 -1 -1 -1 -1 -1 -1
-1 X -1 -1 -1 -1 X X -1
-1 X X -1 -1 X X -1 -1
-1 -1 X 1 -1 1 -1 -1 -1
-1 -1 -1 -1 1 -1 -1 -1 -1
-1 -1 -1 1 -1 1 X -1 -1
-1 -1 X X -1 -1 X X -1
-1 X X -1 -1 -1 -1 X -1
-1 -1 -1 -1 -1 -1 -1 -1 -1
24. 12/13/16Intel Confidential 24
How CNN Works
24
Filtering: The math behind the match
1. Line up the feature and the image patch.
2. Multiply each image pixel by the corresponding feature pixel.
3. Add them up.
4. Divide by the total number of pixels in the feature.
45. 12/13/16Intel Confidential 45
How CNN Works
45
Pooling: Shrinking the image stack
1. Pick a window size (usually 2 or 3).
2. Pick a stride (usually 2).
3. Walk your window across your filtered images.
4. From each window, take the maximum value.