Unsupervised learning of feature hierarchies is often a good initialization for supervised training of deep architectures. In existing deep learning methods, these feature hierarchies are built layer by layer in a greedy fashion using auto-encoders or restricted Boltzmann machines. Both yield encoders, which compute linear projections followed by a smooth thresholding function. In this work, we...

We study two local search algorithms and a greedy algorithm for scheduling. The worstcase performance guarantees are well-known but seem to be contrived and too pessimistic for practical applications. For unrestrictedmachines, Brunsch et al. (2013) showed that the worst-case performance guarantees of these algorithms are not robust if the job sizes are subject to random noise. However, in the c...

We establish the unexpected power of conflict driven clause learning (CDCL) proof search by proving that the sets of unsatisfiable clauses obtained from the guarded graph tautology principles of Alekhnovich, Johannsen, Pitassi and Urquhart have polynomial size pool resolution refutations that use only input lemmas as learned clauses. We further show that, under the correct heuristic choices, th...

We propose a novel, efficient approach for obtaining high-quality experimental designs for event-related functional magnetic resonance imaging (ER-fMRI), a popular brain mapping technique. Our proposed approach combines a greedy hill-climbing algorithm and a cyclic permutation method. When searching for optimal ER-fMRI designs, the proposed approach focuses only on a promising restricted class ...

This paper describes a new heuristic for vehicle routing problems with narrow time windows. The problem arises in the context of the delivery of groceries to restaurants. For most of the instances the given time window distribution does not allow solutions where no time restrictions are violated. The aim is to schedule most of the customers in time building regionally bounded tours. The few rem...