Denver Dash

Non-Boolean associative architectures based on nano-oscillators

Many of the proposed and emerging nano-scale technologies simply cannot compete with CMOS in terms of energy efficiency for performing Boolean operations. However, the potential for these technologies to perform useful non-Boolean computations remains an opportunity to be explored. In this talk we examine the use of the resonance of coupled nano-scale oscillators as a primitive computational operator for associative processing and develop the architectural structures that could enable such devices to be integrated into mainstream applications.

A distributed host-based worm detection system

We present a method for detecting large-scale worm attacks using only end-host detectors. These detectors propagate and aggregate alerts to cooperating partners to detect large-scale distributed attacks in progress. The properties of the host-based detectors may in fact be relatively poor in isolation but when taken collectively result in a high-quality distributed worm detector. We implement a cooperative alert sharing protocol coupled with distributed sequential hypothesis testing to generate global alarms about distributed attacks. We evaluate the systems response in the presence of a variety of false alarm conditions and in the presence of an Internet worm attack. Our evaluation is conducted with agents on the Emulab and DETER emulated testbeds using real operating systems and computing platforms.

Caveats for Causal Reasoning with Equilibrium Models

In this paper we examine the ability to perform causal reasoning with recursive equilibrium models. We identify a critical postulate, which we term the Manipulation Postulate, that is required in order to perform causal inference, and we prove that there exists a general class F of recursive equilibrium models that violate the Manipulation Postulate. We relate this class to the existing phenomenon of reversibility and show that all models in F display reversible behavior, thereby providing an explanation for reversibility and suggesting that it is a special case of a more general and perhaps widespread problem. We also show that all models in F possess a set of variables V′ whose manipulation will cause an instability such that no equilibrium model will exist for the system. We define the Structural Stability Principle which provides a graphical criterion for stability in causal models. Our theorems suggest that drastically incorrect inferences may be obtained when applying the Manipulation Postulate to equilibrium models, a result which has implications for current work on causal modeling, especially causal discovery from data.

Machine learning algorithms for event detection

A common task in many machine learning application domains involves monitoring routinely collected data for ‘interesting’ events. This task is prevalent in surveillance, but also in tasks ranging from the analysis of scientific data to the monitoring of naturally occurring events, and from supervising industrial processes to observing human behavior. We will refer to this monitoring process with the purpose of identifying interesting occurrences, as event detection. We put together this special issue of the Machine Learning journal with the belief that principled machine learning approaches can and will be a differentiator in addressing event detection tasks, and that theoretical and practical advances of machine learning in this area have the potential to impact a wide range of important real-world applications such as security, public health and medicine, biology, environmental sciences, manufacturing, astrophysics, business, and economics. In the recent past, domain experts in these areas have had the laborious job of manually examining the collected data for events of interest. With the emergence of computers, many efforts have been made to replace manual inspection with an automated process. Data, however, have become increasingly complex, and the quantities of collected data have become extremely large in recent years. Multivariate records, images, video footage, audio recordings, spatial and spatio-temporal data, text documents, and even relational data are now routinely collected. We all expect that advances in machine learning would be well-suited for this class of tasks. However, in practice, the peculiarities of the application often grossly violate the

Non-Boolean Associative Processing: Circuits, System Architecture, and Algorithms

We present the design and the performance of a hierarchical associative memory (AM) based on phase locking of coupled oscillators used for pattern recognition. The use of coupled oscillators, rather than Boolean logic, provides for implementations using emerging nanotechnology, such as magnetic spin-torque oscillators and resonant body transistor oscillators, which have the potential of lower energy and higher density than CMOS solutions. We develop a model for the general behavior of weakly coupled nonlinear oscillators that perform pattern matching using a simulation of coupled CMOS ring oscillators. We derive a simple analytic model for their phase locking behavior and use this reduced model in a hierarchical AM for image recognition tasks, such as identifying handwritten numbers.