Parameter Learning for Log-supermodular Distributions

Authors: Tatiana Shpakova, Francis Bach

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In Section 5, we illustrate our new results on a set of experiments in binary image denoising, where we highlight the flexibility of a probabilistic model for learning with missing data. The aim of our experiments is to demonstrate the ability of our approach to remove noise in binary images, following the experimental set-up of [9]. We consider the training sample of Ntrain = 100 images of size D = 50 50, and the test sample of Ntest = 100 binary images, containing a horse silhouette from the Weizmann horse database [3]. Results are presented in Table 1, where we compare the two types of decoding, as well as a structured output SVM (SVM-Struct [22]) applied to the same problem.
Researcher Affiliation Academia Tatiana Shpakova INRIA École Normale Supérieure Paris tatiana.shpakova@inria.fr Francis Bach INRIA École Normale Supérieure Paris francis.bach@inria.fr
Pseudocode Yes Input: functions fk, k = 1, . . . , K, and expected sufficient statistics fk(x) emp. R and x emp. [0, 1]D, regularizer Ω(t, α). Initialization: α = 0, t = 0 Iterations: for h from 1 to H Sample z RD as independent logistics Compute y = y (z, t, α) arg max y {0,1}D z y + t y PK k=1 αkf(y) Replace t by t Ch y x emp. + tΩ(t, α) Replace αk by αk Ch fk(x) emp. fk(y ) + αkΩ(t, α) Output: (α, t).
Open Source Code No No explicit statement providing access to the source code for the methodology described in the paper was found. There are no links to repositories or mentions of code in supplementary materials.
Open Datasets Yes We consider the training sample of Ntrain = 100 images of size D = 50 50, and the test sample of Ntest = 100 binary images, containing a horse silhouette from the Weizmann horse database [3].
Dataset Splits Yes We consider the training sample of Ntrain = 100 images of size D = 50 50, and the test sample of Ntest = 100 binary images, containing a horse silhouette from the Weizmann horse database [3]. one parameter for t, one for α, both learned by cross-validation.
Hardware Specification No No specific hardware details such as CPU/GPU models, memory, or cloud instance types used for running the experiments were provided in the paper.
Software Dependencies No The paper mentions using 'graph-cuts [4]' but does not provide specific version numbers for this or any other software dependencies, which would be necessary for reproducibility.
Experiment Setup Yes We perform parameter inference by maximum likelihood using stochastic subgradient descent (over the logistic samples), with regularization by the squared ℓ2-norm, one parameter for t, one for α, both learned by cross-validation. We apply stochastic subgradient descent for the difference of the two convex functions Alogistic to learn the model parameters and use fixed regularization parameters equal to 10 2.