HAHenry D. I. Abarbanel

Henry D. I. Abarbanel

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Henry D. I. Abarbanel has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Constraining chaos: Enforcing dynamical invariants in the training of reservoir computers

    by Jason A. Platt, Stephen G. Penny, Timothy A. Smith, Tse-Chun Chen, Henry D. I. Abarbanel

    Drawing on ergodic theory, we introduce a novel training method for machine learning based forecasting methods for chaotic dynamical systems. The training enforces dynamical invariants—such as the Lyapunov exponent spectrum and the fractal dimension—in the systems of interest, enabling longer and more stable forecasts when operating with limited data. The technique is demonstrated in detail using reservoir computing, a specific kind of recurrent neural network. Results are given for the Lorenz 1996 chaotic dynamical system and a spectral quasi-geostrophic model of the atmosphere, both typical test cases for numerical weather prediction.


  • Open Access

    Integrating Recurrent Neural Networks With Data Assimilation for Scalable Data‐Driven State Estimation

    by S. G. Penny, T. A. Smith, T.‐C. Chen, J. A. Platt, H.‐Y. Lin, M. Goodliff, H. D. I. Abarbanel

    Data assimilation (DA) is integrated with machine learning in order to perform entirely data‐driven online state estimation. To achieve this, recurrent neural networks (RNNs) are implemented as pretrained surrogate models to replace key components of the DA cycle in numerical weather prediction (NWP), including the conventional numerical forecast model, the forecast error covariance matrix, and the tangent linear and adjoint models. It is shown how these RNNs can be initialized using DA methods to directly update the hidden/reservoir state with observations of the target system. The results indicate that these techniques can be applied to estimate the state of a system for the repeated initialization of short‐term forecasts, even in the absence of a traditional numerical forecast model. Further, it is demonstrated how these integrated RNN‐DA methods can scale to higher dimensions by applying domain localization and parallelization, providing a path for practical applications in NWP.


  • Robust forecasting using predictive generalized synchronization in reservoir computing

    by Jason A. Platt, Adrian Wong, Randall Clark, Stephen G. Penny, Henry D. I. Abarbanel

    Reservoir computers (RCs) are a class of recurrent neural networks (RNNs) that can be used for forecasting the future of observed time series data. As with all RNNs, selecting the hyperparameters in the network to yield excellent forecasting presents a challenge when training on new inputs. We analyze a method based on predictive generalized synchronization (PGS) that gives direction in designing and evaluating the architecture and hyperparameters of an RC. To determine the occurrences of PGS, we rely on the auxiliary method to provide a computationally efficient pre-training test that guides hyperparameter selection. We provide a metric for evaluating the RC using the reproduction of the input system’s Lyapunov exponents that demonstrates robustness in prediction.


  • Open Access

    HVC Interneuron Properties from Statistical Data Assimilation

    by Daniel Breen, Sasha Shirman, Eve Armstrong, Nirag Kadakia, Henry Abarbanel

    Data assimilation (DA) solves the inverse problem of inferring initial conditions given data and a model. Here we use biophysically motivated Hodgkin-Huxley (HH) models of avian HVCI neurons, experimentally obtained recordings of these neurons, and our data assimilation algorithm to infer the full set of parameters and a minimal set of ionic currents precisely reproducing the observed waveform information. We find many distinct validated sets of parameters selected by our DA method and choice of model. We conclude exploring variations on the inverse problem applied to neurons producing accurate or inaccurate results; by manipulating data presented to the algorithm, varying sample rate and waveform; and by manipulating the model by adding and subtracting ionic currents.

  • Open Access

    Neural Circuitry for Recognizing Interspike Interval Sequences

    by Henry D. I. Abarbanel, Sachin S. Talathi

    Sensory systems present environmental information to central nervous system as sequences of action potentials or spikes. How do animals recognize these sequences carrying information about their world? We present a biologically inspired neural circuit designed to enable spike pattern recognition. This circuit is capable of training itself on a given interspike interval (ISI) sequence and is then able to respond to presentations of the same sequence. The essential ingredients of the recognition circuit are (a) a tunable time delay circuit, (b) a spike selection unit, and (c) a tuning mechanism using spike timing dependent plasticity of inhibitory synapses. We have investigated this circuit using Hodgkin-Huxley neuron models connected by realistic excitatory and inhibitory synapses. It is robust in the presence of noise represented as jitter in the spike times of the ISI sequence.


  • Computer simulations of stimulus dependent state switching in basic circuits of bursting neurons

    by Mikhail Rabinovich, Ramón Huerta, Maxim Bazhenov, Alexander K. Kozlov, Henry D. I. Abarbanel

    We investigate the ability of oscillating neural circuits to switch between different states of oscillation in two basic neural circuits. We model two quite distinct small neural circuits. The first circuit is based on invertebrate central pattern generator (CPG) studies [A. I. Selverston and M. Moulins, The Crustacean Stomatogastric System (Springer-Verlag, Berlin, 1987)] and is composed of two neurons coupled via both gap junction and inhibitory synapses. The second consists of coupled pairs of interconnected thalamocortical relay and thalamic reticular neurons with both inhibitory and excitatory synaptic coupling. The latter is an elementary unit of the thalamic networks passing sensory information to the cerebral cortex [M. Steriade, D. A. McCormick, and T. J. Sejnowski, Science 262, 679 (1993)]. Both circuits have contradictory coupling between symmetric parts. The thalamocortical model has excitatory and inhibitory connections and the CPG has reciprocal inhibitory and electrical coupling. We describe the dynamics of the individual neurons in these circuits by conductance based ordinary differential equations of Hodgkin-Huxley type [J. Physiol. (London) 117, 500 (1952)]. Both model circuits exhibit bistability and hysteresis in a wide region of coupling strengths. The two main modes of behavior are in-phase and out-of-phase oscillations of the symmetric parts of the network. We investigate the response of these circuits, while they are operating in bistable regimes, to externally imposed excitatory spike trains with varying interspike timing and small amplitude pulses. These are meant to represent spike trains received by the basic circuits from sensory neurons. Circuits operating in a bistable region are sensitive to the frequency of these excitatory inputs. Frequency variations lead to changes from in-phase to out-of-phase coordination or vice versa. The signaling information contained in a spike train driving the network can place the circuit into one or another state depending on the interspike interval and this happens within a few spikes. These states are maintained by the basic circuit after the input signal is ended. When a new signal of the correct frequency enters the circuit, it can be switched to another state with the same ease.


  • Nonlinear-time-series analysis of chaotic laser dynamics

    by Henry D. I. Abarbanel, Z. Gills, C. Liu, R. Roy

    We report on the analysis of experiments on a neodymium-doped yttrium aluminum garnet laser with an intracavity frequency-doubling crystal. Three modes of the laser were excited in differing polarization configurations. The total intensity of infrared light was observed and then analyzed using methods of nonlinear-time-series analysis. We present clear evidence using global false nearest neighbors that when all polarizations are parallel, the intensity is chaotic with two positive Lyapunov exponents and the system can be embedded in dimension 7. The noise level in this operating condition, which we call type I chaos, is small. When one of the polarizations is perpendicular to the others, the intensity is again chaotic with positive Lyapunov exponents, but there is substantial noise in the signal of high dimensional origin, and no finite embedding dimension appears possible. We call this type II chaos. We suggest that the origin of this phenomenon is the intrinsic quantum noise associated with the generation of green light, which is 25 times more intense in the type II operating configuration than in the first. In past experiments with this system we have found that the type I chaos can be controlled to unstable periodic orbits while type II cannot. In each type of chaotic laser operation we use local false nearest neighbors to demonstrate that the local dimension of the dynamics is 7. This means seven differential equations can capture the full dynamics of these regimes of the laser. We evaluate local and global false nearest neighbors to support our conclusions and determine the Lyapunov spectrum of each type of chaotic behavior. The predictability of type II chaos is shown to be much less than that of type I, and we make local polynomial models in reconstructed-state space to demonstrate that we can predict rather well for type I chaos. Finally we suggest a fairly standard model for the interaction of the infrared light with the nonlinear frequency doubling medium and with a two-level of the active medium. © 1996 The American Physical Society.