From: CLeaR: An adaptive continual learning framework for regression tasks
Symbol | Definition |
---|---|
X | The N×M matrix with N feature samples |
Y | The matrix with N measurements associated with X |
\(\hat {Y}\) | The matrix with N predictions associated with X |
xn | The nth column feature vector, \(\mathbf {x}_{n}^{\mathrm {T}}=X_{n,:}\) |
yn | The nth measurement, yn=Yn,: |
\(\hat {y}_{n}\) | The nth prediction, \(\hat {y}_{n} = \hat {Y}_{n,:}\) |
D | The dataset, \(D=\{\left (\mathbf {x}_{n}, y_{n} \right) | n=1, \dots N \}\) |
Θl | The weight matrix of the lth layer, \(\Theta _{l} \in \mathbb {R}^{{l-1} \times {l}}\) |
\(\theta ^{i}_{l}\) | The ith element of Θl |
fl | The activation function of the lth layer, \(f_{l}\colon \mathbb {R} \to \mathbb {R}\) |
fΘ(·) | The neural network with given weight martrix Θ |
zl | The output column vector of the lth layer |
L | Loss function |
P | Probability density |
\(\mathcal {N}(\mu, \sigma ^{2})\) | Gaussian distribution with mean μ and variance σ2 |
Ft | The Fisher information matrix of the tth task |
\(F_{t}^{i}\) | The ith diagonal element of Ft |