This is an introductory workshop on Deep Learning with Keras. We start from shallow models: Linear Regression and Logistic Regression and show how they can be implemented with Keras. Then we show how to move to deeper models, how to use more complex architectures and layers.
The workshop explores common use cases and suggests next steps to apply Keras to solve your problems.
34. Catalit LLC
BINARY CLASSIFICATION
Age Gender
Annual
Salary
Months in
residence
Months in
job
Current
Debt
Paid off
credit
Client 1 23 M $30,000 36 12 $5,000 Yes
Client 2 30 F $45,000 12 12 $1,000 Yes
Client 3 19 M $15,000 3 1 $10,000 No
Features Labels
Data Point
40. Catalit LLC
CONFUSION MATRIX
• Accuracy: Overall, how often is it correct?
• (TP +TN) / total
Test Negative Test Positive
Condition
Negative
TRUE NEGATIVE
FALSE POSITIVE
(Type I error)
Condition
Positive
FALSE NEGATIVE
(Type II error)
TRUE POSITIVE
41. Catalit LLC
CLASSIFICATION SCORES
• Accuracy: Overall, how often correct?
• (TP +TN) / total
• Precision: Test positive, how often prediction correct?
• TP / test yes
• Recall: Actual value positive, how often prediction correct?
• TP / actual yes
Test Negative Test Positive
Condition
Negative
TRUE NEGATIVE
FALSE POSITIVE
(Type I error)
Condition
Positive
FALSE NEGATIVE
(Type II error)
TRUE POSITIVE
80. Catalit LLC
INPUTTENSOR
Input: order 4 tensor
(N, H, W, C)
(60000, 28, 28, 1)
MNIST training set
N: Number of images
H: Height of image
W: Width of image
C: Number of color channels