Showing papers by "Hava T. Siegelmann published in 2019"

Locally Connected Spiking Neural Networks for Unsupervised Feature Learning

Abstract: In recent years, Spiking Neural Networks (SNNs) have demonstrated great successes in completing various Machine Learning tasks. We introduce a method for learning image features by \textit{locally connected layers} in SNNs using spike-timing-dependent plasticity (STDP) rule. In our approach, sub-networks compete via competitive inhibitory interactions to learn features from different locations of the input space. These \textit{Locally-Connected SNNs} (LC-SNNs) manifest key topological features of the spatial interaction of biological neurons. We explore biologically inspired n-gram classification approach allowing parallel processing over various patches of the the image space. We report the classification accuracy of simple two-layer LC-SNNs on two image datasets, which match the state-of-art performance and are the first results to date. LC-SNNs have the advantage of fast convergence to a dataset representation, and they require fewer learnable parameters than other SNN approaches with unsupervised learning. Robustness tests demonstrate that LC-SNNs exhibit graceful degradation of performance despite the random deletion of large amounts of synapses and neurons.

Lattice Map Spiking Neural Networks (LM-SNNs) for Clustering and Classifying Image Data

Abstract: Spiking neural networks (SNNs) with a lattice architecture are introduced in this work, combining several desirable properties of SNNs and self-organized maps (SOMs). Networks are trained with biologically motivated, unsupervised learning rules to obtain a self-organized grid of filters via cooperative and competitive excitatory-inhibitory interactions. Several inhibition strategies are developed and tested, such as (i) incrementally increasing inhibition level over the course of network training, and (ii) switching the inhibition level from low to high (two-level) after an initial training segment. During the labeling phase, the spiking activity generated by data with known labels is used to assign neurons to categories of data, which are then used to evaluate the network's classification ability on a held-out set of test data. Several biologically plausible evaluation rules are proposed and compared, including a population-level confidence rating, and an $n$-gram inspired method. The effectiveness of the proposed self-organized learning mechanism is tested using the MNIST benchmark dataset, as well as using images produced by playing the Atari Breakout game.

Improved robustness of reinforcement learning policies upon conversion to spiking neuronal network platforms applied to ATARI games

Abstract: Deep Reinforcement Learning (RL) demonstrates excellent performance on tasks that can be solved by trained policy. It plays a dominant role among cutting-edge machine learning approaches using multi-layer Neural networks (NNs). At the same time, Deep RL suffers from high sensitivity to noisy, incomplete, and misleading input data. Following biological intuition, we involve Spiking Neural Networks (SNNs) to address some deficiencies of deep RL solutions. Previous studies in image classification domain demonstrated that standard NNs (with ReLU nonlinearity) trained using supervised learning can be converted to SNNs with negligible deterioration in performance. In this paper, we extend those conversion results to the domain of Q-Learning NNs trained using RL. We provide a proof of principle of the conversion of standard NN to SNN. In addition, we show that the SNN has improved robustness to occlusion in the input image. Finally, we introduce results with converting full-scale Deep Q-network to SNN, paving the way for future research to robust Deep RL applications.

Minibatch Processing in Spiking Neural Networks

Abstract: Spiking neural networks (SNNs) are a promising candidate for biologically-inspired and energy efficient computation. However, their simulation is notoriously time consuming, and may be seen as a bottleneck in developing competitive training methods with potential deployment on neuromorphic hardware platforms. To address this issue, we provide an implementation of mini-batch processing applied to clock-based SNN simulation, leading to drastically increased data throughput. To our knowledge, this is the first general-purpose implementation of mini-batch processing in a spiking neural networks simulator, which works with arbitrary neuron and synapse models. We demonstrate nearly constant-time scaling with batch size on a simulation setup (up to GPU memory limits), and showcase the effectiveness of large batch sizes in two SNN application domains, resulting in $\approx$880X and $\approx$24X reductions in wall-clock time respectively. Different parameter reduction techniques are shown to produce different learning outcomes in a simulation of networks trained with spike-timing-dependent plasticity. Machine learning practitioners and biological modelers alike may benefit from the drastically reduced simulation time and increased iteration speed this method enables. Code to reproduce the benchmarks and experimental findings in this paper can be found at this https URL.

Abstraction Mechanisms Predict Generalization in Deep Neural Networks

Abstract: A longstanding problem for Deep Neural Networks (DNNs) is understanding their puzzling ability to generalize well. We approach this problem through the unconventional angle of \textit{cognitive abstraction mechanisms}, drawing inspiration from recent neuroscience work, allowing us to define the Cognitive Neural Activation metric (CNA) for DNNs, which is the correlation between information complexity (entropy) of given input and the concentration of higher activation values in deeper layers of the network. The CNA is highly predictive of generalization ability, outperforming norm-and-margin-based generalization metrics on an extensive evaluation of over 100 dataset-and-network-architecture combinations, especially in cases where additive noise is present and/or training labels are corrupted. These strong empirical results show the usefulness of CNA as a generalization metric, and encourage further research on the connection between information complexity and representations in the deeper layers of networks in order to better understand the generalization capabilities of DNNs.

Relating information complexity and training in deep neural networks

Abstract: Deep Neural Networks may be costly to train, and if testing error is too large, retraining may be required, unless lifelong learning methods are applied. Crucial to addressing learning at the edge, without access to powerful cloud computing, is the notion of problem difficulty for non-standard data domains. While it is known that training is harder for classes that are more entangled, the complexity of data points was not previously studied as an important contributor to training dynamics. We analyze data points by their information complexity and relate the complexity of the data to the test error. We elucidate training dynamics of DNNs, demonstrating that high complexity datapoints contribute to the error of the network, and that training DNNs consist of two important aspects - (1) Minimization of error due to high complexity datapoints, and (2) Margin decrease where entanglement of classes occurs. Whereas data complexity may be ignored when training in a cloud, it must be considered as part of the setting when training at the edge.

Gibbs Free Energy, a thermodynamic measure of protein-protein interactions, correlates with neurologic disability

Abstract: Background: Modern network science has been used to reveal new and often fundamental aspects of brain network organization in physiological as well as pathological conditions. As a consequence, these discoveries, which relate to network hierarchy, hubs and network interactions, begun to change the paradigms of neurodegenerative disorders. We therefore explored the use of thermodynamics for protein-protein network interactions in Alzheimer disease (AD), Parkinson disease (PD), multiple sclerosis (MS), traumatic brain injury and epilepsy. Methods: To assess the validity of using network interactions in neurological disease, we investigated the relationship between network thermodynamics and molecular systems biology for these neurological disorders. In order to uncover whether there was a correlation between network organization and biological outcomes, we used publicly available RNA transcription data from individual patients with these neurological conditions, and correlated these molecular profiles with their respective individual disability scores. Results: We found a linear correlation (Pearson correlation of -0.828) between disease disability (a clinically validated measurement of a person9s functional status), and Gibbs free energy (a thermodynamic measure of protein-protein interactions). In other words, we found an inverse relationship between disease entropy and thermodynamic energy. Interpretation: Because a larger degree of disability correlated with a larger negative drop in Gibbs free energy in a linear, disability-dependent fashion, it could be presumed that the progression of neuropathology such as is seen in Alzheimer Disease, could potentially be prevented by therapeutically correcting the changes Gibbs free energy.

Models of Situated Intelligence Inspired by the Energy Management of Brains

Abstract: Embodiment is a key feature of biological intelligence, and energy-awareness can be viewed as the ultimate expression of situated intelligence. Energy constraint is often ignored, or it has just secondary role in typical cutting-edge AI approaches. For example, Deep Learning Networks often require huge amount of data/ time/ energy/ resources, which may not be readily available in various practical scenarios. We outline a simple computational model of spiking neural activity lined link astrocytes responsible for energy management in brains. We analyze oscillatory neural dynamics and its modulation by astrocytes mediating energy constrains. We indicate the potential benefits of the proposed design to be implemented on neuromorphic computational platforms.

Thermodynamic Measures of Human Brain Development from Fetal Stage to Adulthood

Abstract: This paper analyzes the data obtained from tissue samples of the human brains containing protein expression values. The data have been processed for their thermodynamic measure in terms of the Gibbs free energy of the corresponding protein-protein interaction networks. We have investigated the functional dependence of the Gibbs free energies on age and found consistent trends for most of the 16 main brain areas. The peak of the Gibbs energy values is found at birth with a trend toward plateauing at the age of maturity. We have also compared the data for males and females and uncovered functional differences for some of the brain regions.

Deep Neural Networks Abstract Like Humans.

Abstract: Deep neural networks (DNNs) have revolutionized AI due to their remarkable performance in pattern recognition, comprising of both memorizing complex training sets and demonstrating intelligence by generalizing to previously unseen data (test sets). The high generalization performance in DNNs has been explained by several mathematical tools, including optimization, information theory, and resilience analysis. In humans, it is the ability to abstract concepts from examples that facilitates generalization; this paper thus researches DNN generalization from that perspective. A recent computational neuroscience study revealed a correlation between abstraction and particular neural firing patterns. We express these brain patterns in a closed-form mathematical expression, termed the `Cognitive Neural Activation metric' (CNA) and apply it to DNNs. Our findings reveal parallels in the mechanism underlying abstraction in DNNs and those in the human brain. Beyond simply measuring similarity to human abstraction, the CNA is able to predict and rate how well a DNN will perform on test sets, and determines the best network architectures for a given task in a manner not possible with extant tools. These results were validated on a broad range of datasets (including ImageNet and random labeled datasets) and neural architectures.

Utilizing full neuronal states for adversarial robustness

Abstract: Small, imperceptible perturbations of the input data can lead to DNNs making egregious errors during inference, such as misclassifying an image of a dog as a cat with high probability. Thus, defending against adversarial examples for deep neural networks (DNNs) is of great interest to sensing technologies and the machine learning community to ensure the security of practical systems where DNNs are used. Whereas many approaches have been explored for defending against adversarial attacks, few have made use of the full state of the entire network, opting instead to only consider the output layer and gradient information. We develop several motivated techniques that make use of the full network state, improving adversarial robustness. We provide principled motivation of our techniques via analysis of attractor dynamics, shown to occur in the highly recurrent human brain, and validate our improvements via empirical results on standard datasets and white-box attacks.