The traditional approach to distributed machine learning is adapt algorithms the network, e.g., reducing updates curb overhead. Networks based on intelligent edge, instead, make it possible follow opposite approach, i.e., define logical network topology around task perform, so as meet desired performance. In this paper, we propose a system model that captures such aspects in context of supervis...

Binary networks are extremely efficient as they use only two symbols to define the network: {+1, −1}. One can make prior distribution of these a design choice. The recent IR-Net Qin et al. argues that imposing Bernoulli with equal priors (equal bit ratios) over binary weights leads maximum entropy and thus minimizes information loss. However, work cannot precisely control weight during training...

IP-based networks form the base of todays communication infrastructure. The interconnection of formerly isolated networks brings up severe security issues. The standard approach, to protect the own network from abuse, is the usage of filter mechanisms at the border to the foreign network. The raising complexity of protocols and the use of encryption techniques render most of these borderoriente...

We propose a new paradigm for network defense: a hierarchical network of lightweight, mobile and adaptive tools combined with a distributed, collaborative intrusion detection environment. Agents are integrated with the collaborative IDS in order to provide them with a wider array of information to use in their response activities. This provides for both a mobile and adaptive defense, while stil...