HSHava Tova Siegelmann

Hava Tova Siegelmann

*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


Hava Tova Siegelmann has a total of 111 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Optimizing Multilinear Polynomial Representations of Boolean Functions for Neural Network Synthesis

    by Joshua Russell, Ignacio Gavier, Edward Rietman, Hava Siegelmann

    We consider the problem of obtaining a neural network (NN) representation of a Boolean network (BN), where each Boolean network (BN) vertex is a k -input Boolean function (BF) ( k≤ K ). A neural network (NN) representation can be constructed by mapping each Boolean function (BF) to its multilinear polynomial (MP) representation and then to a two-layer sub-NN; composing sub-NNs completes the representation. In this article, we show that structural properties (number of neurons/connections, neuron in-degree) of such a NN representation can be efficiently optimized while maintaining functional equivalence with the original BN ( lossless optimization). Our approach establishes a correspondence between multilinear polynomials (MPs) and their two-layer-NN representation, enabling sub-NN optimization by minimizing MPs under structural criteria. To accelerate optimization, we utilize criterion invariance under Negation-Permutation-Negation (NPN) transformations. This allows us to classify BFs into equivalence classes, optimize canonical representatives, and then derive optimal representations for remaining class members via Negation-Permutation-Negation (NPN) transformations. Since optimizing sub-NNs may affect the composed NN depth, we propose an algorithm for lossless optimization under depth constraints. Experimentally, across four control-oriented and cryptographic digital circuits, we achieve reductions of up to 59% in neurons and 69% in connections relative to non-optimized NNs. Under minimum-depth constraints, reductions of up to 23% in neurons and 50% in connections are obtained. We demonstrate minimization of mean and maximum in-degree, and that intermediate-depth solutions recover most of the size reduction. Our NPN-based acceleration reduces MP optimization time by up to 10.2× compared to a baseline caching solutions per BF. Overall, the proposed methods facilitate high-throughput BN simulation and neurosymbolic matrix-vector-multiplication architectures.

  • Open Access

    Turing universal neural networks do not require global clocks

    by Hava T. Siegelmann, Roy N. Siegelmann, Stephen Chung, Chloé Becquey

    Recurrent neural networks were proven to be Turing universal in the 1990s, motivating computational complexity studies of spiking networks, neural Turing machines with differentiable activations, and transformers. At the time, neural networks were exploratory and small, whereas today large-scale deployment makes energy efficiency critical. We thus extend the development of computational foundations of neural networks to asynchronous networks. Asynchrony is modeled by updating a single randomly selected neuron per step, eliminating global updates and reducing energy use. While asynchrony introduces variability in update sequences and thus has often been considered impractical for computing, we introduce design constraints which lead to Turing universal asynchronous architectures. We prove universality both for asynchronous fixed architectures with varying-precision neurons and for variable architectures with fixed-precision neurons. These results advance the theoretical understanding of asynchronous networks, suggesting that they preserve full computational power, remain amenable for efficient training, and may achieve substantial reductions in energy use.

  • Open Access

    Neuromodulators Generate Multiple Context-Relevant Behaviors in Recurrent Neural Networks

    by Ben Tsuda, Stefan C. Pate, Kay M. Tye, Hava T. Siegelmann, Terrence J. Sejnowski

    Neuromodulators are critical controllers of neural states, with dysfunctions linked to various neuropsychiatric disorders. Although many biological aspects of neuromodulation have been studied, the computational principles underlying how neuromodulation of distributed neural populations controls brain states remain unclear. In contrast to external contextual inputs, neuromodulation can act as a single scalar signal that is broadcast to a vast population of neurons. We model the modulation of synaptic weight in a recurrent neural network model and show that neuromodulators can dramatically alter the function of a network, even when highly simplified. We find that under structural constraints like those in brains, this provides a fundamental mechanism that can increase the computational capability and flexibility of a neural network. Diffuse synaptic weight modulation enables storage of multiple memories using a common set of synapses that are able to generate diverse, even diametrically opposed, behaviors. Our findings help explain how neuromodulators unlock specific behaviors by creating task-specific hyperchannels in neural activity space and motivate more flexible, compact and capable machine learning architectures.


  • Open Access

    Exponential Dynamic Energy Network for High Capacity Sequence Memory

    by Arjun Karuvally, Pichsinee Lertsaroj, Terrence J. Sejnowski, Hava T. Siegelmann

    The energy paradigm, exemplified by Hopfield networks, offers a principled framework for memory in neural systems by interpreting dynamics as descent on an energy surface. While powerful for static associative memories, it falls short in modeling sequential memory, where transitions between memories are essential. We introduce the Exponential Dynamic Energy Network (EDEN), a novel architecture that extends the energy paradigm to temporal domains by evolving the energy function over multiple timescales. EDEN combines a static high-capacity energy network with a slow, asymmetrically interacting modulatory population, enabling robust and controlled memory transitions. We formally derive short-timescale energy functions that govern local dynamics and use them to analytically compute memory escape times, revealing a phase transition between static and dynamic regimes. The analysis of capacity, defined as the number of memories that can be stored with minimal error rate as a function of the dimensions of the state space (number of feature neurons), for EDEN shows that it achieves exponential sequence memory capacity $O(γ^N)$, outperforming the linear capacity $O(N)$ of conventional models. Furthermore, EDEN's dynamics resemble the activity of time and ramping cells observed in the human brain during episodic memory tasks, grounding its biological relevance. By unifying static and sequential memory within a dynamic energy framework, EDEN offers a scalable and interpretable model for high-capacity temporal memory in both artificial and biological systems.

  • Open Access

    Reservoir Computation with Networks of Differentiating Neuron Ring Oscillators

    by Alexander Yeung, Peter DelMastro, Arjun Karuvally, Hava Siegelmann, Edward Rietman, Hananel Hazan

    Reservoir computing is an approach to machine learning that leverages the dynamics of a complex system alongside a simple, often linear, machine learning model for a designated task. While many efforts have previously focused their attention on integrating neurons, which produce an output in response to large, sustained inputs, we focus on using differentiating neurons, which produce an output in response to large changes in input. Here, we introduce a small-world graph built from rings of differentiating neurons as a Reservoir Computing substrate. We find the coupling strength and network topology that enable these small-world networks to function as an effective reservoir. The dynamics of differentiating neurons naturally give rise to oscillatory dynamics when arranged in rings, where we study their computational use in the Reservoir Computing setting. We demonstrate the efficacy of these networks in the MNIST digit recognition task, achieving comparable performance of 90.65% to existing Reservoir Computing approaches. Beyond accuracy, we conduct systematic analysis of our reservoir’s internal dynamics using three complementary complexity measures that quantify neuronal activity balance, input dependence, and effective dimensionality. Our analysis reveals that optimal performance emerges when the reservoir operates with intermediate levels of neural entropy and input sensitivity, consistent with the edge-of-chaos hypothesis, where the system balances stability and responsiveness. The findings suggest that differentiating neurons can be a potential alternative to integrating neurons and can provide a sustainable future alternative for power-hungry AI applications.


  • Open Access

    Finger joint angle and gesture estimation under natural conditions with a soft printed electrode array

    by Nitzan Luxembourg, Rufael Fekadu Marew, Dvir Teitelbaum, Dvir Ben-Dov, Hava Siegelmann, Yael Hanein

    Innovative methods for finger gesture recognition have been an active research area, with surface electromyography (sEMG) emerging as a promising approach in human-machine interface applications, especially when visual imaging is impractical. However, sEMG-based gesture recognition is highly susceptible to movement artifacts, individual muscle activation, and changes in hand position, making dynamic gesture recognition challenging. While progress has been made in sEMG data collection and analysis, most studies focus on controlled, static hand positions, limiting real-world applicability. This study integrates a soft wearable sEMG sensor, a Video-Vision-Transform model, and motion sensor-based training to predict finger joint angles and recognize gestures across both static and dynamic hand positions. Despite inter-subject variability, results demonstrate differentiation of finger angles and gestures. For the highly performing subject, recognition accuracy reached 0.85 for static and 0.87 for dynamic settings. This work advances sEMG-based gesture recognition, indicating stable performance across tested static and dynamic conditions, suggesting potential suitability for natural and real-world applications.


  • Open Access

    Reservoir Computation with Networks of Differentiating Neuron Ring Oscillators

    by Alexander Yeung, Peter DelMastro, Arjun Karuvally, Hava Siegelmann, Edward Rietman, Hananel Hazan

    Reservoir Computing is a machine learning approach that uses the rich repertoire of complex system dynamics for function approximation. Current approaches to reservoir computing use a network of coupled integrating neurons that require a steady current to maintain activity. Here, we introduce a small world graph of differentiating neurons that are active only when there are changes in input as an alternative to integrating neurons as a reservoir computing substrate. We find the coupling strength and network topology that enable these small world networks to function as an effective reservoir. We demonstrate the efficacy of these networks in the MNIST digit recognition task, achieving comparable performance of 90.65% to existing reservoir computing approaches. The findings suggest that differentiating neurons can be a potential alternative to integrating neurons and can provide a sustainable future alternative for power-hungry AI applications.

  • Open Access

    Transient Dynamics in Lattices of Differentiating Ring Oscillators

    by Peter DelMastro, Arjun Karuvally, Hananel Hazan, Hava Siegelmann, Edward Rietman

    Recurrent neural networks (RNNs) are machine learning models widely used for learning temporal relationships. Current state-of-the-art RNNs use integrating or spiking neurons -- two classes of computing units whose outputs depend directly on their internal states -- and accordingly there is a wealth of literature characterizing the behavior of large networks built from these neurons. On the other hand, past research on differentiating neurons, whose outputs are computed from the derivatives of their internal states, remains limited to small hand-designed networks with fewer than one-hundred neurons. Here we show via numerical simulation that large lattices of differentiating neuron rings exhibit local neural synchronization behavior found in the Kuramoto model of interacting oscillators. We begin by characterizing the periodic orbits of uncoupled rings, herein called ring oscillators. We then show the emergence of local correlations between oscillators that grow over time when these rings are coupled together into lattices. As the correlation length grows, transient dynamics arise in which large regions of the lattice settle to the same periodic orbit, and thin domain boundaries separate adjacent, out-of-phase regions. The steady-state scale of these correlated regions depends on how the neurons are shared between adjacent rings, which suggests that lattices of differentiating ring oscillator might be tuned to be used as reservoir computers. Coupled with their simple circuit design and potential for low-power consumption, differentiating neural nets therefore represent a promising substrate for neuromorphic computing that will enable low-power AI applications.

  • Open Access

    Distributed Multi-Agent Lifelong Learning

    by Prithviraj Tarale, Edward Rietman, Hava Siegelmann

    Lifelong learning (LL) systems must adapt to changing environments by continuously updating their knowledge. Traditional LL paradigms assume new data come labeled and that agents learn independently. However, labeled data are scarce, especially in remote settings. We introduce the Peer Parallel Lifelong Learning (PEEPLL) framework for distributed Multi-Agent LL, where agents actively seek peer assistance when needed. Unlike classical distributed AI, where communication scales poorly, PEEPLL agents reduce communication as their knowledge evolves. To improve resilience to low-quality peer responses, we propose (a) TRUE confidence score, a compute-efficient application of Variational Autoencoder without input reconstruction, (b) REFINE algorithm to selectively accept peer responses, (c) DYMEM, Dynamic Memory-Update for storing necessary information. Each of the three contributions improves upon their baselines. In our solution to PEEPLL, agents outperform traditional LL agents, even when the latter has environmental supervision available, marking a substantial step towards scalable lifelong learning at the edge.


  • Optimizing Attention and Cognitive Control Costs Using Temporally Layered Architectures

    by Devdhar Patel, Terrence Sejnowski, Hava Siegelmann

    The current reinforcement learning framework focuses exclusively on performance, often at the expense of efficiency. In contrast, biological control achieves remarkable performance while also optimizing computational energy expenditure and decision frequency. We propose a decision-bounded Markov decision process (DB-MDP) that constrains the number of decisions and computational energy available to agents in reinforcement learning environments. Our experiments demonstrate that existing reinforcement learning algorithms struggle within this framework, leading to either failure or suboptimal performance. To address this, we introduce a biologically inspired, temporally layered architecture (TLA), enabling agents to manage computational costs through two layers with distinct timescales and energy requirements. TLA achieves optimal performance in decision-bounded environments and in continuous control environments, matching state-of-the-art performance while using a fraction of the computing cost. Compared to current reinforcement learning algorithms that solely prioritize performance, our approach significantly lowers computational energy expenditure while maintaining performance. These findings establish a benchmark and pave the way for future research on energy and time-aware control.