Learning the Valuations of a $k$-demand Agent

Authors: Hanrui Zhang, Vincent Conitzer

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In Section 4, we experimentally evaluate the performance of ERM algorithms.
Researcher Affiliation Academia 1Department of Computer Science, Duke University, Durham, USA. Correspondence to: Hanrui Zhang <hrzhang@cs.duke.edu>, Vincent Conitzer <conitzer@cs.duke.edu>.
Pseudocode Yes Algorithm 1 Biased Binary Search
Open Source Code No The paper does not contain any statement about making its source code publicly available, nor does it provide a link to a code repository.
Open Datasets No We draw the ground truth value vector uniformly at random from the unit hypercube [0, 1]n, and for each sample, we draw the price vector uniformly at random from [−1, 0]n. The paper generates its own data for experiments and does not use a pre-existing public dataset with concrete access information.
Dataset Splits No The paper mentions training and testing but does not explicitly describe a validation dataset split.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using an 'LP solver' but does not provide specific version numbers for it or any other software dependencies.
Experiment Setup Yes We implement the ERM learner by solving the system in Proposition 2 using an LP solver, where the objective is to maximize the minimum margin. We draw the ground truth value vector uniformly at random from the unit hypercube [0, 1]n, and for each sample, we draw the price vector uniformly at random from [−1, 0]n. To study the performance of ERM for different values of k, we fix the number of items to be n = 50, and examine the accuracy of the ERM learner for k ∈ {5, 10, 15, 20, 25} respectively. To study the performance of ERM for different values of n, we fix the agent to be unit-demand (i.e., k = 1), and calculate the accuracy of the ERM learner for n ∈ {20, 40, 60, 80, 100} respectively. In both experiments, we let the size of the training set grow, and plot the empirical error rate for each size of the training set ℓ ∈ {50, 100, 150, 200, 250, 300, 350, 400, 450, 500}.