University of Southern California

Bart Kosko

Osonde Osoba

Kartik Audhkhasi

Sanya Mitaim

Brandon Franzke

Peter J Pacini

A learning computer system may estimate unknown parameters and states of a stochastic or uncertain system having a probability structure. The system may include a data processing system that may include a hardware processor that has a configuration that: receives data; generates random, chaotic, fuzzy, or other numerical perturbations of the data, one or more of the states, or the probability structure; estimates observed and hidden states of the stochastic or uncertain system using the data, the generated perturbations, previous states of the stochastic or uncertain system, or estimated states of the stochastic or uncertain system; and causes perturbations or independent noise to be injected into the data, the states, or the stochastic or uncertain system so as to speed up training or learning of the probability structure and of the system parameters or the states.

Noise speed-ups in hidden markov models with applications to speech recognition

The invention shows how to use noise-like perturbations to improve the speed and accuracy of Markov Chain Monte Carlo (MCMC) estimates and large-scale optimization, simulated annealing optimization, and quantum annealing for large-scale optimization.

Using noise to speed convergence of simulated annealing and Markov Monte Carlo estimations

A learning computer system may include a data processing system and a hardware processor and may estimate parameters and states of a stochastic or uncertain system. The system may receive data from a user or other source. Parameters and states of the stochastic or uncertain system are estimated using the received data, numerical perturbations, and previous parameters and states of the stochastic or uncertain system. It is determined whether the generated numerical perturbations satisfy a condition. If the numerical perturbations satisfy the condition, the numerical perturbations are injected into the estimated parameters or states, the received data, the processed data, the masked or filtered data, or the processing units.

Noise-enhanced convolutional neural networks

An estimating computer system may iteratively estimate an unknown parameter of a model or state of a system. An input module may receive numerical data about the system. A noise module may generate random, chaotic, or other type of numerical perturbations of the received numerical data and/or may generate pseudo-random noise. An estimation module may iteratively estimate the unknown parameter of the model or state of the system based on the received numerical data. The estimation module may use the numerical perturbations and/or the pseudo-random noise and the input numerical data during at least one of the iterative estimates of the unknown parameter. A signaling module may signal when successive parameter estimates or information derived from successive parameter estimates differ by less than a predetermined signaling threshold or when the number of estimation iterations reaches a predetermined number.

Iterative estimation of system parameters using noise-like perturbations

A system and method for fuzzy spread spectrum communication. The system includes a fuzzy spreader for spreading an input signal over a range of frequencies and fuzzy despreader for extracting the spread input signal from the range of frequencies. In the illustrative implementation, the inventive system a fuzzy pseudo-random generator for use in the fuzzy spreader and the fuzzy despreader. The fuzzy pseudo-random generator uses a novel method for generating pseudo-random numbers. It does not use encryption or decryption techniques. The invention further provides a method for adaptive rule generation and a novel method for identifying the centroid of the set of output numbers. This allows the fuzzy system to learn spreading rules that favor data compression, compact multiplexing, bandwidth conservation, and other communication tasks as well as rules that favor security.

Growing community of inventors

Hacienda Heights, CA, United States of America

Bart Kosko