Professional Documents
Culture Documents
Today data driven modules are necessary in products to achieve the next level of performance. Using
Honeywell use cases this session will help the audience gain an understanding of the concepts and tools
for building data driven intelligent systems. This session will cover:
3. Classification
a. Training a Binary Classifier
b. Performance Measures
i. Measuring Accuracy Using Cross-Validation
ii. Confusion Matrix
iii. Precision and Recall
iv. Precision/Recall Tradeoff
v. The ROC Curve
c. Multiclass Classification
d. Error Analysis
e. Multilabel Classification
f. Multioutput Classification
4. Training Models
a. Linear Regression
i. The Normal Equation
ii. Computational Complexity
b. Gradient Descent
i. Batch Gradient Descent
ii. Stochastic Gradient Descent
iii. Mini-batch Gradient Descent
c. Polynomial Regression
d. Learning Curves
e. Regularized Linear Models
i. Ridge Regression
ii. Lasso Regression
iii. Elastic Net
iv. Early Stopping
f. Logistic Regression
i. Estimating Probabilities
ii. Training and Cost Function
iii. Decision Boundaries
iv. Softmax Regression
6. Decision Trees
a. Training and Visualizing a Decision Tree
b. Making Predictions
c. Estimating Class Probabilities
d. The CART Training Algorithm
e. Computational Complexity
f. Gini Impurity or Entropy?
g. Regularization Hyperparameters
h. Regression
i. Instability
8. Dimensionality Reduction
a. Main Approaches for Dimensionality Reduction
i. Projection
ii. Manifold Learning
b. PCA
i. Preserving the Variance
ii. Principal Components
iii. Projecting Down to d Dimensions
iv. Explained Variance Ratio
v. Choosing the Right Number of Dimensions
vi. PCA for Compression
vii. Randomized PCA
viii. Incremental PCA
c. Kernel PCA
i. Selecting a Kernel and Tuning Hyperparameters