We study the problem of minimizing the diameter of a graph by adding k shortcut edges, for speeding up communication in an existing network design. We develop constant-factor approximation algorithms for different variations of this problem. We also show how to improve the approximation ratios using resource augmentation to allow more than k shortcut edges. We observe a close relation between t...

It is well accepted that convolutional neural networks play an important role in learning excellent features for image classification and recognition. However, in tradition they only allow adjacent layers connected, limiting integration of multi-scale information. To further improve their performance, we present a concatenating framework of shortcut convolutional neural networks. This framework...

Shortcut connections are a popular architectural feature of multi-layer perceptrons. It is generally assumed that by implementing a linear submapping, shortcuts assist the learning process in the remainder of the network. Here we find that this is not always the case: shortcut weights may also act as distractors that slow down convergence and can lead to inferior solutions. This problem can be ...

We propose a technique for mining minimum sets with bounded reachability in real-world networks i.e.~the smallest vertex sets such that any other vertex is at distance at most $k$ from at least one vertex of this set. Our technique uses a simple but efficient mechanism. We first introduce new edges to shorten the paths in our network to obtain a shortcut graph. As the next step, we search for t...

Deep stacked RNNs are usually hard to train. Recent studies have shown that shortcut connections across different RNN layers bring substantially faster convergence. However, shortcuts increase the computational complexity of the recurrent computations. To reduce the complexity, we propose the shortcut block, which is a refinement of the shortcut LSTM blocks. Our approach is to replace the self-...

Shortcut fusion is a well-known optimization technique for functional programs. Its aim is to transform multi-pass algorithms into single pass ones, achieving deforestation of the intermediate structures that multi-pass algorithms need to construct. Shortcut fusion has already been extended in several ways. It can be applied to monadic programs, maintaining the global effects, and also to obtai...

In functional programming one usually writes programs as the composition of simpler functions. Consequently, the result of a function might be generated only to be consumed immediately by another function. This potential source of inefficiency can often be eliminated using a technique called shortcut fusion, which fuses both functions involved in a composition to yield a monolithic one. In this...

One crucial feature of expertise is the ability to spontaneously recognize where and when knowledge can be applied to simplify task processing. Mental arithmetic is one domain in which people should start to develop such expert knowledge in primary school by integrating conceptual knowledge about mathematical principles and procedural knowledge about shortcuts. If successful, knowledge integrat...

An existential optimal landmark is a set of actions, one of which must be used in some optimal plan. Recently, Karpas and Domshlak (2012) introduced a technique for deriving such existential optimal landmarks, which is based on using shortcut rules — rules which take a path, and attempt to find a cheaper path that achieves some of the propositions that the original path achieved. The shortcut r...