Panoptic Diffusion Models: co-generation of images and segmentation maps

Recently, diffusion models have demonstrated impressive capabilities in text-guided and image-conditioned image generation. However, existing diffusion models cannot simultaneously generate a segmentation map of objects and a corresponding image from the prompt. Previous attempts either generate segmentation maps based on the images or provide maps as input conditions to control image generation, limiting their functionality to given inputs. Incorporating an inherent understanding of the scene layouts can improve the creativity and realism of diffusion models. To address this limitation, we present Panoptic Diffusion Model (PDM), the first model designed to generate both images and panoptic segmentation maps concurrently. PDM bridges the gap between image and text by constructing segmentation layouts that provide detailed, built-in guidance throughout the generation process. This ensures the inclusion of categories mentioned in text prompts and enriches the diversity of segments within the background. We demonstrate the effectiveness of PDM across two architectures: a unified diffusion transformer and a two-stream transformer with a pretrained backbone. To facilitate co-generation with fewer sampling steps, we incorporate a fast diffusion solver into PDM. Additionally, when ground-truth maps are available, PDM can function as a text-guided image-to-image generation model. Finally, we propose a novel metric for evaluating the quality of generated maps and show that PDM achieves state-of-the-art results in image generation with implicit scene control.

TSkips: Efficiency Through Explicit Temporal Delay Connections in Spiking Neural Networks

Spiking Neural Networks (SNNs) with their bio-inspired Leaky Integrate-and-Fire (LIF) neurons inherently capture temporal information. This makes them well-suited for sequential tasks like processing event-based data from Dynamic Vision Sensors (DVS) and event-based speech tasks. Harnessing the temporal capabilities of SNNs requires mitigating vanishing spikes during training, capturing spatio-temporal patterns and enhancing precise spike timing. To address these challenges, we propose TSkips, augmenting SNN architectures with forward and backward skip connections that incorporate explicit temporal delays. These connections capture long-term spatio-temporal dependencies and facilitate better spike flow over long sequences. The introduction of TSkips creates a vast search space of possible configurations, encompassing skip positions and time delay values. To efficiently navigate this search space, this work leverages training-free Neural Architecture Search (NAS) to identify optimal network structures and corresponding delays. We demonstrate the effectiveness of our approach on four event-based datasets: DSEC-flow for optical flow estimation, DVS128 Gesture for hand gesture recognition and Spiking Heidelberg Digits (SHD) and Spiking Speech Commands (SSC) for speech recognition. Our method achieves significant improvements across these datasets: up to 18% reduction in Average Endpoint Error (AEE) on DSEC-flow, 8% increase in classification accuracy on DVS128 Gesture, and up to 8% and 16% higher classification accuracy on SHD and SSC, respectively.

CODE-CL: COnceptor-Based Gradient Projection for DEep Continual Learning

Continual learning, or the ability to progressively integrate new concepts, is fundamental to intelligent beings, enabling adaptability in dynamic environments. In contrast, artificial deep neural networks face the challenge of catastrophic forgetting when learning new tasks sequentially. To alleviate the problem of forgetting, recent approaches aim to preserve essential weight subspaces for previous tasks by limiting updates to orthogonal subspaces via gradient projection. While effective, this approach can lead to suboptimal performance, particularly when tasks are highly correlated. In this work, we introduce COnceptor-based gradient projection for DEep Continual Learning (CODE-CL), a novel method that leverages conceptor matrix representations, a computational model inspired by neuroscience, to more flexibly handle highly correlated tasks. CODE-CL encodes directional importance within the input space of past tasks, allowing new knowledge integration in directions modulated by $1-S$, where $S$ represents the direction's relevance for prior tasks. Additionally, we analyze task overlap using conceptor-based representations to identify highly correlated tasks, facilitating efficient forward knowledge transfer through scaled projection within their intersecting subspace. This strategy enhances flexibility, allowing learning in correlated tasks without significantly disrupting previous knowledge. Extensive experiments on continual learning image classification benchmarks validate CODE-CL's efficacy, showcasing superior performance with minimal forgetting, outperforming most state-of-the-art methods.

SpiDR: A Reconfigurable Digital Compute-in-Memory Spiking Neural Network Accelerator for Event-based Perception

Spiking Neural Networks (SNNs), with their inherent recurrence, offer an efficient method for processing the asynchronous temporal data generated by Dynamic Vision Sensors (DVS), making them well-suited for event-based vision applications. However, existing SNN accelerators suffer from limitations in adaptability to diverse neuron models, bit precisions and network sizes, inefficient membrane potential (Vmem) handling, and limited sparse optimizations. In response to these challenges, we propose a scalable and reconfigurable digital compute-in-memory (CIM) SNN accelerator \chipname with a set of key features: 1) It uses in-memory computations and reconfigurable operating modes to minimize data movement associated with weight and Vmem data structures while efficiently adapting to different workloads. 2) It supports multiple weight/Vmem bit precision values, enabling a trade-off between accuracy and energy efficiency and enhancing adaptability to diverse application demands. 3) A zero-skipping mechanism for sparse inputs significantly reduces energy usage by leveraging the inherent sparsity of spikes without introducing high overheads for low sparsity. 4) Finally, the asynchronous handshaking mechanism maintains the computational efficiency of the pipeline for variable execution times of different computation units. We fabricated \chipname in 65 nm Taiwan Semiconductor Manufacturing Company (TSMC) low-power (LP) technology. It demonstrates competitive performance (scaled to the same technology node) to other digital SNN accelerators proposed in the recent literature and supports advanced reconfigurability. It achieves up to 5 TOPS/W energy efficiency at 95% input sparsity with 4-bit weights and 7-bit Vmem precision.

Power side-channel leakage localization through adversarial training of deep neural networks

Supervised deep learning has emerged as an effective tool for carrying out power side-channel attacks on cryptographic implementations. While increasingly-powerful deep learning-based attacks are regularly published, comparatively-little work has gone into using deep learning to defend against these attacks. In this work we propose a technique for identifying which timesteps in a power trace are responsible for leaking a cryptographic key, through an adversarial game between a deep learning-based side-channel attacker which seeks to classify a sensitive variable from the power traces recorded during encryption, and a trainable noise generator which seeks to thwart this attack by introducing a minimal amount of noise into the power traces. We demonstrate on synthetic datasets that our method can outperform existing techniques in the presence of common countermeasures such as Boolean masking and trace desynchronization. Results on real datasets are weak because the technique is highly sensitive to hyperparameters and early-stop point, and we lack a holdout dataset with ground truth knowledge of leaking points for model selection. Nonetheless, we believe our work represents an important first step towards deep side-channel leakage localization without relying on strong assumptions about the implementation or the nature of its leakage. An open-source PyTorch implementation of our experiments is provided.

ASMA: An Adaptive Safety Margin Algorithm for Vision-Language Drone Navigation via Scene-Aware Control Barrier Functions

In the rapidly evolving field of vision-language navigation (VLN), ensuring robust safety mechanisms remains an open challenge. Control barrier functions (CBFs) are efficient tools which guarantee safety by solving an optimal control problem. In this work, we consider the case of a teleoperated drone in a VLN setting, and add safety features by formulating a novel scene-aware CBF using ego-centric observations obtained through an RGB-D sensor. As a baseline, we implement a vision-language understanding module which uses the contrastive language image pretraining (CLIP) model to query about a user-specified (in natural language) landmark. Using the YOLO (You Only Look Once) object detector, the CLIP model is queried for verifying the cropped landmark, triggering downstream navigation. To improve navigation safety of the baseline, we propose ASMA -- an Adaptive Safety Margin Algorithm -- that crops the drone's depth map for tracking moving object(s) to perform scene-aware CBF evaluation on-the-fly. By identifying potential risky observations from the scene, ASMA enables real-time adaptation to unpredictable environmental conditions, ensuring optimal safety bounds on a VLN-powered drone actions. Using the robot operating system (ROS) middleware on a parrot bebop2 quadrotor in the gazebo environment, ASMA offers 59.4% - 61.8% increase in success rates with insignificant 5.4% - 8.2% increases in trajectory lengths compared to the baseline CBF-less VLN while recovering from unsafe situations.

SHIRE: Enhancing Sample Efficiency using Human Intuition in REinforcement Learning

The ability of neural networks to perform robotic perception and control tasks such as depth and optical flow estimation, simultaneous localization and mapping (SLAM), and automatic control has led to their widespread adoption in recent years. Deep Reinforcement Learning has been used extensively in these settings, as it does not have the unsustainable training costs associated with supervised learning. However, DeepRL suffers from poor sample efficiency, i.e., it requires a large number of environmental interactions to converge to an acceptable solution. Modern RL algorithms such as Deep Q Learning and Soft Actor-Critic attempt to remedy this shortcoming but can not provide the explainability required in applications such as autonomous robotics. Humans intuitively understand the long-time-horizon sequential tasks common in robotics. Properly using such intuition can make RL policies more explainable while enhancing their sample efficiency. In this work, we propose SHIRE, a novel framework for encoding human intuition using Probabilistic Graphical Models (PGMs) and using it in the Deep RL training pipeline to enhance sample efficiency. Our framework achieves 25-78% sample efficiency gains across the environments we evaluate at negligible overhead cost. Additionally, by teaching RL agents the encoded elementary behavior, SHIRE enhances policy explainability. A real-world demonstration further highlights the efficacy of policies trained using our framework.

DCT-CryptoNets: Scaling Private Inference in the Frequency Domain

The convergence of fully homomorphic encryption (FHE) and machine learning offers unprecedented opportunities for private inference of sensitive data. FHE enables computation directly on encrypted data, safeguarding the entire machine learning pipeline, including data and model confidentiality. However, existing FHE-based implementations for deep neural networks face significant challenges in computational cost, latency, and scalability, limiting their practical deployment. This paper introduces DCT-CryptoNets, a novel approach that leverages frequency-domain learning to tackle these issues. Our method operates directly in the frequency domain, utilizing the discrete cosine transform (DCT) commonly employed in JPEG compression. This approach is inherently compatible with remote computing services, where images are usually transmitted and stored in compressed formats. DCT-CryptoNets reduces the computational burden of homomorphic operations by focusing on perceptually relevant low-frequency components. This is demonstrated by substantial latency reduction of up to 5.3$\times$ compared to prior work on image classification tasks, including a novel demonstration of ImageNet inference within 2.5 hours, down from 12.5 hours compared to prior work on equivalent compute resources. Moreover, DCT-CryptoNets improves the reliability of encrypted accuracy by reducing variability (e.g., from $\pm$2.5\% to $\pm$1.0\% on ImageNet). This study demonstrates a promising avenue for achieving efficient and practical privacy-preserving deep learning on high resolution images seen in real-world applications.

Approximate ADCs for In-Memory Computing

In memory computing (IMC) architectures for deep learning (DL) accelerators leverage energy-efficient and highly parallel matrix vector multiplication (MVM) operations, implemented directly in memory arrays. Such IMC designs have been explored based on CMOS as well as emerging non-volatile memory (NVM) technologies like RRAM. IMC architectures generally involve a large number of cores consisting of memory arrays, storing the trained weights of the DL model. Peripheral units like DACs and ADCs are also used for applying inputs and reading out the output values. Recently reported designs reveal that the ADCs required for reading out the MVM results, consume more than 85% of the total compute power and also dominate the area, thereby eschewing the benefits of the IMC scheme. Mitigation of imperfections in the ADCs, namely, non-linearity and variations, incur significant design overheads, due to dedicated calibration units. In this work we present peripheral aware design of IMC cores, to mitigate such overheads. It involves incorporating the non-idealities of ADCs in the training of the DL models, along with that of the memory units. The proposed approach applies equally well to both current mode as well as charge mode MVM operations demonstrated in recent years., and can significantly simplify the design of mixed-signal IMC units.

Eigen Attention: Attention in Low-Rank Space for KV Cache Compression

Large language models (LLMs) represent a groundbreaking advancement in the domain of natural language processing due to their impressive reasoning abilities. Recently, there has been considerable interest in increasing the context lengths for these models to enhance their applicability to complex tasks. However, at long context lengths and large batch sizes, the key-value (KV) cache, which stores the attention keys and values, emerges as the new bottleneck in memory usage during inference. To address this, we propose Eigen Attention, which performs the attention operation in a low-rank space, thereby reducing the KV cache memory overhead. Our proposed approach is orthogonal to existing KV cache compression techniques and can be used synergistically with them. Through extensive experiments over OPT, MPT, and Llama model families, we demonstrate that Eigen Attention results in up to 40% reduction in KV cache sizes and up to 60% reduction in attention operation latency with minimal drop in performance.