Incremental Randomized Sketching for Online Kernel Learning

Authors: Xiao Zhang, Shizhong Liao

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate that the incremental randomized sketching achieves a better learning performance in terms of accuracy and efficiency even in adversarial environments.
Researcher Affiliation Academia 1College of Intelligence and Computing, Tianjin University, Tianjin 300350, China.
Pseudocode Yes Algorithm 1 Ske GD Algorithm
Open Source Code No The paper does not provide concrete access to source code for the methodology described, nor does it contain an explicit code release statement or repository link.
Open Datasets Yes We compare our Ske GD with the state-of-the-art online kernel learning algorithms on the well-known classification benchmarks4. [...] 4https://www.csie.ntu.edu.tw/~cjlin/libsvm
Dataset Splits No The paper mentions merging 'training and testing data into a single dataset' and performing '20 different random permutations', but it does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology for train/validation/test sets).
Hardware Specification Yes All experiments are performed on a machine with 4-core Intel Core i7 3.60 GHz CPU and 16GB memory.
Software Dependencies Yes The compared algorithms are obtained from the LIBOL v0.3.0 toolbox and LSOKL toolbox5.
Experiment Setup Yes The stepsizes η of all the gradient descent based algorithms are tuned in 10[ 5:+1:0], and the regularization parameters λ are tuned in 10[ 4:+1:1]. Besides, we use the Gaussian kernel κ(x, x ) = exp x x 2 2/2σ2 , where the set σ {2[ 5:+0.5:7]} are adopted as the candidate kernel set. [...] Besides, we set τ = 0.2, sp = 3B/4, sm = τsp, d = 4 and ρ = 0.3T in our Ske GD if not specially specified, and the rank k = 0.1B for NOGD and Ske GD.