I have 20 years experience designing and implementing machine learning algorithms for data analysis applied to physical systems. I would characterize myself as a Computational Physicist, although I have done a great deal of both experimental and theoretical work.My areas of expertise include Astrophysics, Bayesian Probability Theory, Information Theory, Robotics, Signal Processing, Neuroscience, Nonlinear Dynamics, Quantum Mechanics, Foundations of Physics and Earth Science.
American Astronomical Society Meeting Abstracts, Dec 1, 2006
ABSTRACT The Aromatic Infrared Bands (AIBs) are prominent features of many galactic spectra in th... more ABSTRACT The Aromatic Infrared Bands (AIBs) are prominent features of many galactic spectra in the infrared and are strongly associated with star-forming regions. Polycyclic Aromatic Hydrocarbons (PAHs) are a leading contender as the source for this emission. Unequivocally establishing this identification has been difficult since the AIBs do not appear to be the result of emission from a small, tractable number of PAHs. Rather the observed emission bands appear to be composed of the contributions from possibly hundreds of different PAHs, neutral and ionized, each with its own distinctive spectrum. A major step toward verifying the PAH hypothesis would be to quantitatively establish how well the observed AIBs can be explained, or not explained, as the combination of known PAH spectra. To date this problem has been attacked by either manually superimposing individual PAH spectra or by using non-negative least squares. We explain how both of these approaches have serious deficiencies. We then describe our progress in applying Bayesian source separation techniques to this difficult problem. In this phase of our investigation, we have worked with theoretically generated PAH spectra at a single temperature composed of combinations of arbitrarily selected PAHs with added Gaussian noise. First, we show how the non-negative least squares approach fares as a function of PAH composition, noise level, and spectral resolution. Next, we outline our Bayesian approach, which relies on Skilling's nested sampling algorithm. This approach allows us to find highly probable solutions and evaluate the uncertainties in our estimates by sampling the posterior. Moreover, this algorithm also enables us to compute the evidence provided by the data as well as visualize the posterior probability in the high-dimensional hypothesis space generated by the numerous combinations of possible PAH contributions. This research is supported by NASA Applied Information Systems Research Grant 05-AISR05-0143.
ABSTRACT Daisyworld was first introduced in 1983 by Watson and Lovelock as a model that illustrat... more ABSTRACT Daisyworld was first introduced in 1983 by Watson and Lovelock as a model that illustrates how life can influence a planet's climate. These models typically involve modeling a planetary surface on which black and white daisies can grow thus influencing the local surface albedo and therefore also the temperature distribution. Since then, variations of daisyworld have been applied to study problems ranging from ecological systems to global climate. Much of the interest in daisyworld models is due to the fact that they enable one to study self-regulating systems. These models are nonlinear, and as such they exhibit sensitive dependence on initial conditions, and depending on the specifics of the model they can also exhibit feedback loops, oscillations, and chaotic behavior. Many daisyworld models are thermodynamic in nature in that they rely on heat flux and temperature gradients. However, what is not well-known is whether, or even why, a daisyworld model might settle into a maximum entropy production (MEP) state. With the aim to better understand these systems, this paper will discuss what is known about the role of MEP in daisyworld models.
Neural activity in the auditory cortex of the human brain can be evoked by presenting transient a... more Neural activity in the auditory cortex of the human brain can be evoked by presenting transient acoustic stimuli to the ear. Using the Magnes Biomagnetometer System at the Scripps Clinic in La Jolla CA, we measured the magnetic fields generated by this neural activity. The auditory steady-state response (SSR), first studied by Galambos et al.1, is a sinusoidal response elicited by periodic trains of auditory click transients presented at rates ranging from 15–70 Hz. Galambos et al. proposed that the SSR was generated by a superposition of the individual transient middle-latency responses.
Studies confirming the influence of visual information on the response of the auditory cortex to ... more Studies confirming the influence of visual information on the response of the auditory cortex to speech sounds [1] lead us to question the nature of the connection between the visual and auditory cortices. The superior colliculus exhibits visual-sensitive neurons that can be modulated (both inhibited and excited) by auditory stimuli [2]. Since the superior colliculus is one of the nuclei that the optic nerve passes through as it travels to the visual cortex, this suggests that one might expect a modulation of the visual response with a simultaneous auditory stimuli.
American Astronomical Society Meeting Abstracts, Dec 1, 2006
ABSTRACT The Aromatic Infrared Bands (AIBs) are prominent features of many galactic spectra in th... more ABSTRACT The Aromatic Infrared Bands (AIBs) are prominent features of many galactic spectra in the infrared and are strongly associated with star-forming regions. Polycyclic Aromatic Hydrocarbons (PAHs) are a leading contender as the source for this emission. Unequivocally establishing this identification has been difficult since the AIBs do not appear to be the result of emission from a small, tractable number of PAHs. Rather the observed emission bands appear to be composed of the contributions from possibly hundreds of different PAHs, neutral and ionized, each with its own distinctive spectrum. A major step toward verifying the PAH hypothesis would be to quantitatively establish how well the observed AIBs can be explained, or not explained, as the combination of known PAH spectra. To date this problem has been attacked by either manually superimposing individual PAH spectra or by using non-negative least squares. We explain how both of these approaches have serious deficiencies. We then describe our progress in applying Bayesian source separation techniques to this difficult problem. In this phase of our investigation, we have worked with theoretically generated PAH spectra at a single temperature composed of combinations of arbitrarily selected PAHs with added Gaussian noise. First, we show how the non-negative least squares approach fares as a function of PAH composition, noise level, and spectral resolution. Next, we outline our Bayesian approach, which relies on Skilling's nested sampling algorithm. This approach allows us to find highly probable solutions and evaluate the uncertainties in our estimates by sampling the posterior. Moreover, this algorithm also enables us to compute the evidence provided by the data as well as visualize the posterior probability in the high-dimensional hypothesis space generated by the numerous combinations of possible PAH contributions. This research is supported by NASA Applied Information Systems Research Grant 05-AISR05-0143.
ABSTRACT Daisyworld was first introduced in 1983 by Watson and Lovelock as a model that illustrat... more ABSTRACT Daisyworld was first introduced in 1983 by Watson and Lovelock as a model that illustrates how life can influence a planet's climate. These models typically involve modeling a planetary surface on which black and white daisies can grow thus influencing the local surface albedo and therefore also the temperature distribution. Since then, variations of daisyworld have been applied to study problems ranging from ecological systems to global climate. Much of the interest in daisyworld models is due to the fact that they enable one to study self-regulating systems. These models are nonlinear, and as such they exhibit sensitive dependence on initial conditions, and depending on the specifics of the model they can also exhibit feedback loops, oscillations, and chaotic behavior. Many daisyworld models are thermodynamic in nature in that they rely on heat flux and temperature gradients. However, what is not well-known is whether, or even why, a daisyworld model might settle into a maximum entropy production (MEP) state. With the aim to better understand these systems, this paper will discuss what is known about the role of MEP in daisyworld models.
Neural activity in the auditory cortex of the human brain can be evoked by presenting transient a... more Neural activity in the auditory cortex of the human brain can be evoked by presenting transient acoustic stimuli to the ear. Using the Magnes Biomagnetometer System at the Scripps Clinic in La Jolla CA, we measured the magnetic fields generated by this neural activity. The auditory steady-state response (SSR), first studied by Galambos et al.1, is a sinusoidal response elicited by periodic trains of auditory click transients presented at rates ranging from 15–70 Hz. Galambos et al. proposed that the SSR was generated by a superposition of the individual transient middle-latency responses.
Studies confirming the influence of visual information on the response of the auditory cortex to ... more Studies confirming the influence of visual information on the response of the auditory cortex to speech sounds [1] lead us to question the nature of the connection between the visual and auditory cortices. The superior colliculus exhibits visual-sensitive neurons that can be modulated (both inhibited and excited) by auditory stimuli [2]. Since the superior colliculus is one of the nuclei that the optic nerve passes through as it travels to the visual cortex, this suggests that one might expect a modulation of the visual response with a simultaneous auditory stimuli.
Uploads