Ayon Borthakur

Vaishnavi Nagabhushana, Kartikay Agrawal, Ayon Borthakur

Although deep neural networks perform extremely well in controlled environments, they fail in real-world scenarios where data isn't available all at once, and the model must adapt to a new data distribution that may or may not follow the initial distribution. Previously acquired knowledge is lost during subsequent updates based on new data. a phenomenon commonly known as catastrophic forgetting. In contrast, the brain can learn without such catastrophic forgetting, irrespective of the number of tasks it encounters. Existing spiking neural networks (SNNs) for class-incremental learning (CIL) suffer a sharp performance drop as tasks accumulate. We here introduce CATFormer (Context Adaptive Threshold Transformer), a scalable framework that overcomes this limitation. We observe that the key to preventing forgetting in SNNs lies not only in synaptic plasticity but also in modulating neuronal excitability. At the core of CATFormer is the Dynamic Threshold Leaky Integrate-and-Fire (DTLIF) neuron model, which leverages context-adaptive thresholds as the primary mechanism for knowledge retention. This is paired with a Gated Dynamic Head Selection (G-DHS) mechanism for task-agnostic inference. Extensive evaluation on both static (CIFAR-10/100/Tiny-ImageNet) and neuromorphic (CIFAR10-DVS/SHD) datasets reveals that CATFormer outperforms existing rehearsal-free CIL algorithms across various task splits, establishing it as an ideal architecture for energy-efficient, true-class incremental learning.

Om Govind Jha, Manoj Bamniya, Ayon Borthakur

Traditional discriminative computer vision relies predominantly on static projections, mapping input features to outputs in a single computational step. Although efficient, this paradigm lacks the iterative refinement and robustness inherent in biological vision and modern generative modelling. In this paper, we propose Discriminative Flow Matching, a framework that reformulates classification and object detection as a conditional transport process. By learning a vector field that continuously transports samples from a simple noise distribution toward a task-aligned target manifold -- such as class embeddings or bounding box coordinates -- we are at the interface between generative and discriminative learning. Our method attaches multiple independent flow predictors to a shared backbone. These predictors are trained using local flow matching objectives, where gradients are computed independently for each block. We formulate this approach for standard image classification and extend it to the complex task of object detection, where targets are high-dimensional and spatially distributed. This architecture provides the flexibility to update blocks either sequentially to minimise activation memory or in parallel to suit different hardware constraints. By aggregating the predictions from these independent flow predictors, our framework enables robust, generative-inspired inference across diverse architectures, including CNNs and vision transformers.

Aditya Somasundaram, Pushkal Mishra, Ayon Borthakur

We propose a learning algorithm to overcome the limitations of traditional backpropagation in resource-constrained environments: Solo Pass Embedded Learning Algorithm (SPELA). SPELA operates with local loss functions to update weights, significantly saving on resources allocated to the propagation of gradients and storing computational graphs while being sufficiently accurate. Consequently, SPELA can closely match backpropagation using less memory. Moreover, SPELA can effectively fine-tune pre-trained image recognition models for new tasks. Further, SPELA is extended with significant modifications to train CNN networks, which we evaluate on CIFAR-10, CIFAR-100, and SVHN 10 datasets, showing equivalent performance compared to backpropagation. Our results indicate that SPELA, with its features such as local learning and early exit, is a potential candidate for learning in resource-constrained edge AI applications.

Kartikay Agrawal, Abhijeet Vikram, Vedant Sharma et al.

In recent years, with the emergence of large models, there has been a significant interest in spiking neural networks (SNNs) primarily due to their energy efficiency, multiplication-free, and sparse event-based deep learning. Similarly, state space models (SSMs) in varying designs have evolved as a powerful alternative to transformers for target modeling in long sequences, thereby overcoming the quadratic dependence on sequence length of a transformer. Inspired by this progress, we here design SHaRe-SSM (Spiking Harmonic Resonate and Fire State Space Model), for target variable modeling (including both classification and regression) for very-long-range sequences. Our second-order spiking SSM, on average, performs better than transformers or first-order SSMs while circumventing multiplication operations, making it ideal for resource-constrained applications. The proposed block consumes $73 \times$ less energy than second-order ANN-based SSMs for an 18k sequence, while retaining performance. To ensure learnability over the long-range sequences, we propose exploiting the stable and efficient implementation of the dynamical system using parallel scans. Moreover, for the first time, we propose a kernel-based spiking regressor using resonate and fire neurons for very long-range sequences. Our network shows superior performance on even a 50k sequence while being significantly energy-efficient. In addition, we conducted a systematic analysis of the impact of heterogeneity, dissipation, and conservation in resonate-and-fire SSMs.

Ayon Borthakur, Thomas A. Cleland

The mammalian olfactory system learns rapidly from very few examples, presented in unpredictable online sequences, and then recognizes these learned odors under conditions of substantial interference without exhibiting catastrophic forgetting. We have developed a brain-mimetic algorithm that replicates these properties, provided that sensory inputs adhere to a common statistical structure. However, in natural, unregulated environments, this constraint cannot be assured. We here present a series of signal conditioning steps, inspired by the mammalian olfactory system, that transform diverse sensory inputs into a regularized statistical structure to which the learning network can be tuned. This pre-processing enables a single instantiated network to be applied to widely diverse classification tasks and datasets - here including gas sensor data, remote sensing from spectral characteristics, and multi-label hierarchical identification of wild species - without adjusting network hyperparameters.