Learning rate
Experts

1

2

3

4

5

6

7

8

9

10
Publications (19)

1A Comparison and Improvement of Online Learning Algorithms for Sequence Labeling2012  Zhengyan He,Houfeng Wangfeature function that maps sequence to feature vector ηt learning rate for the tth sample λ regularization weight of λ 2 ‖θ‖22...

2Accelerated Training of Maximum Margin Markov Models for Sequence Labeling: A Case Study of NP Chunking2010  Xiaofeng Yu,Wai LamMonroe, 1951): wt+1 ← wt − η · ∇wL (15) where η is the learning rate in the algorithm. The SGD algorithm has been shown to...

3Robust Learning in Random Subspaces: Equipping NLP for OOV Effects2012  Anders Søgaard,Anders Johannsenw with a weight for each feature, a bias term b and a learning rate α. For a data point x j , c(x j) = 1 iff w · x + b > 0...

4Resolving Surface Forms to Wikipedia Topics)()( 1 , where j is the node that xi belongs to, r is learning rate. 3: End for 4: Return fM 1337 In our setting,...

5A Comparative Study on Ranking and Selection Strategies for MultiDocument Summarizationupdated in a gradient descent manner: w=w η∆w and η is the learning rate. For details, refer to (Cao et al., 2007). 4 ILPbased...

6Inducing Crosslingual Distributed Representations of Wordsform: cw ← cw +η∂ L (t)(θ) ∂ cw , (4) where η is the learning rate and L(t)(θ) = log Pˆθ (wt wt−n+1:t−1) is the contribution...

7Learning Semantics with Deep Belief Network for CrossLanguage Information Retrievalupdated by learning rate 0.1, weight decay 2× 10−4 and momentum 0.9 except the first RBMs, in which the learning rate was set...

8Unsupervised Discriminative Induction of Synchronous Grammar for Machine Translationsource hypergraph 8 λ← λ + η × ∂ L ∂ λ (H1, H2) ⊲ η is learning rate 9 return G′, λ and store them in G′. After that, we...

9Improvements to Training an RNN parsergradient of Ex for each training example, multiplied by a learning rate. 281 Stacking logistic regression layers into neural...

10Factored Language Model based on Recurrent Neural Networkis the learning rate. After each iteration, it uses validation data for stopping and controlling the learning rate. Usually...