Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Nonasymptotic convergence of stochastic proximal point methods for constrained convex optimization
Authors: Andrei Patrascu, Ion Necoara
JMLR 2017 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical evidence supports the effectiveness of our methods in real problems. (...) We present numerical evidence to assess the theoretical convergence guarantees of the SPP algorithm. We provide three numerical examples: constrained stochastic least-square with random generated data (...) Markowitz portfolio optimization using real data (...) and logistic regression using real data (...). In all our figures the results are averaged over 20 Monte-Carlo simulations for an algorithm. |
| Researcher Affiliation | Academia | Andrei Patrascu EMAIL Department of Computer Science University of Bucharest Str. Academiei 14, 010014 Bucharest Ion Necoara EMAIL Automatic Control and Systems Engineering Department University Politehnica of Bucharest Spl. Independentei 313, 060042 Bucharest |
| Pseudocode | Yes | Algorithm SPP (x0, {µk}k 0) (...) Algorithm RSPP Let µ0 > 0 and x0,0 Rn. For t 1 do: 1. Compute stepsize µt and number of inner iterations Kt 2. Set x Kt,t the average output of SPP(x Kt 1,t 1, µt) runned for Kt iterations with constant stepsize µt 3. If an outer stopping criterion is satisfied, then STOP, otherwise t := t+1 and go to step 1. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code for the methodology described, nor does it provide a link to a code repository. |
| Open Datasets | Yes | We use 2 different real portfolio datasets: Standard & Poor s 500 (SP500, with 25 stocks for 1276 days) and one dataset by Fama and French (FF100, with 100 portfolios for 23.647 days) that is commonly used in financial literature, see e.g. (Brodie et al., 2009). (...) We have tested the four schemes (SGD, SPP, ASPP and RSPP), on the Adult datasets (a2a with m = 2265, n = 123 and a5a with m = 6414, n = 123) from LIBSVM/UCI database (Platt, 1998). |
| Dataset Splits | Yes | We split all the datasets into test (10%) and train (90%) partitions randomly. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run its experiments, such as CPU/GPU models or memory specifications. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software dependencies used in the experiments. It describes the algorithms but does not mention the implementation environment or libraries with versions. |
| Experiment Setup | Yes | In Figure 1 we run algorithms SPP, RSPP, A-SPP and SGD for two values of the initial stepsize: µ0 = 0.5 and µ0 = 1. Each scheme runs for two stepsize exponents: γ1 = 1 (left) and γ2 = 1/2 (right). (...) We set the initial stepsize at value µ0 = 0.6 and the regularization parameter λ = 10 3. |