Convolutional Layer: a matrix (also viewed as a test function with compact support) with filters (pattern detection)
1st layer: applied convolution between the test func. and raw data (matrix) → output data is also a matrix
An Introduction to Gradient Boosting Decision Trees - Machine Learning Plus
例：第1個回歸模型使用原始資料，其後第N+1個回歸模型利用第N個回歸模型產生的殘差。