A domain-agnostic approach for characterization of lifelong learning systems.

Despite the advancement of machine learning techniques in recent years, state-of-the-art systems lack robustness to "real world" events, where the input distributions and tasks encountered by the deployed systems will not be limited to the original training context, and systems will instead need to adapt to novel distributions and tasks while deployed. This critical gap may be addressed through the development of "Lifelong Learning" systems that are capable of (1) Continuous Learning, (2) Transfer and Adaptation, and (3) Scalability. Unfortunately, efforts to improve these capabilities are typically treated as distinct areas of research that are assessed independently, without regard to the impact of each separate capability on other aspects of the system. We instead propose a holistic approach, using a suite of metrics and an evaluation framework to assess Lifelong Learning in a principled way that is agnostic to specific domains or system techniques. Through five case studies, we show that this suite of metrics can inform the development of varied and complex Lifelong Learning systems. We highlight how the proposed suite of metrics quantifies performance trade-offs present during Lifelong Learning system development - both the widely discussed Stability-Plasticity dilemma and the newly proposed relationship between Sample Efficient and Robust Learning. Further, we make recommendations for the formulation and use of metrics to guide the continuing development of Lifelong Learning systems and assess their progress in the future.

Exploring Neuromodulation for Dynamic Learning.

A continual learning system requires the ability to dynamically adapt and generalize to new tasks with access to only a few samples. In the central nervous system, across species, it is observed that continual and dynamic behavior in learning is an active result of a mechanism known as neuromodulation. Therefore, in this work, neuromodulatory plasticity is embedded with dynamic learning architectures as a first step toward realizing power and area efficient few shot learning systems. An inbuilt modulatory unit regulates learning based on the context and internal state of the system. This renders the system an ability to self modify its weights. In one of the proposed architectures, ModNet, a modulatory layer is introduced in a random projection framework. ModNet's learning capabilities are enhanced by integrating attention along with compartmentalized plasticity mechanisms. Moreover, to explore modulatory mechanisms in conjunction with backpropagation in deeper networks, a modulatory trace learning rule is introduced. The proposed learning rule, uses a time dependent trace to modify the synaptic connections as a function of ongoing states and activations. The trace itself is updated via simple plasticity rules thus reducing the demand on resources. The proposed ModNet and learning rules demonstrate the ability to learn from few samples, train quickly, and perform few-shot image classification in a computationally efficient manner. The simple ModNet and the compartmentalized ModNet architecture learn benchmark image classification tasks in just 2 epochs. The network with modulatory trace achieves an average accuracy of 98.8%±1.16 on the omniglot dataset for five-way one-shot image classification task while requiring 20x fewer trainable parameters in comparison to other state of the art models.

Deep Liquid State Machines With Neural Plasticity for Video Activity Recognition.

Real-world applications such as first-person video activity recognition require intelligent edge devices. However, size, weight, and power constraints of the embedded platforms cannot support resource intensive state-of-the-art algorithms. Machine learning lite algorithms, such as reservoir computing, with shallow 3-layer networks are computationally frugal as only the output layer is trained. By reducing network depth and plasticity, reservoir computing minimizes computational power and complexity, making the algorithms optimal for edge devices. However, as a trade-off for their frugal nature, reservoir computing sacrifices computational power compared to state-of-the-art methods. A good compromise between reservoir computing and fully supervised networks are the proposed deep-LSM networks. The deep-LSM is a deep spiking neural network which captures dynamic information over multiple time-scales with a combination of randomly connected layers and unsupervised layers. The deep-LSM processes the captured dynamic information through an attention modulated readout layer to perform classification. We demonstrate that the deep-LSM achieves an average of 84.78% accuracy on the DogCentric video activity recognition task, beating state-of-the-art. The deep-LSM also shows up to 91.13% memory savings and up to 91.55% reduction in synaptic operations when compared to similar recurrent neural network models. Based on these results we claim that the deep-LSM is capable of overcoming limitations of traditional reservoir computing, while maintaining the low computational cost associated with reservoir computing.

HTM Spatial Pooler With Memristor Crossbar Circuits for Sparse Biometric Recognition.

Hierarchical Temporal Memory (HTM) is an online machine learning algorithm that emulates the neo-cortex. The development of a scalable on-chip HTM architecture is an open research area. The two core substructures of HTM are spatial pooler and temporal memory. In this work, we propose a new Spatial Pooler circuit design with parallel memristive crossbar arrays for the 2D columns. The proposed design was validated on two different benchmark datasets, face recognition, and speech recognition. The circuits are simulated and analyzed using a practical memristor device model and 0.18 µm IBM CMOS technology model. The databases AR, YALE, ORL, and UFI, are used to test the performance of the design in face recognition. TIMIT dataset is used for the speech recognition.

Design and Analysis of a Neuromemristive Reservoir Computing Architecture for Biosignal Processing.

Reservoir computing (RC) is gaining traction in several signal processing domains, owing to its non-linear stateful computation, spatiotemporal encoding, and reduced training complexity over recurrent neural networks (RNNs). Previous studies have shown the effectiveness of software-based RCs for a wide spectrum of applications. A parallel body of work indicates that realizing RNN architectures using custom integrated circuits and reconfigurable hardware platforms yields significant improvements in power and latency. In this research, we propose a neuromemristive RC architecture, with doubly twisted toroidal structure, that is validated for biosignal processing applications. We exploit the device mismatch to implement the random weight distributions within the reservoir and propose mixed-signal subthreshold circuits for energy efficiency. A comprehensive analysis is performed to compare the efficiency of the neuromemristive RC architecture in both digital(reconfigurable) and subthreshold mixed-signal realizations. Both Electroencephalogram (EEG) and Electromyogram (EMG) biosignal benchmarks are used for validating the RC designs. The proposed RC architecture demonstrated an accuracy of 90 and 84% for epileptic seizure detection and EMG prosthetic finger control, respectively.