Training at the edge utilizes continuously evolving data generated at different locations. Privacy concerns prohibit the co-location of this spatially as well as temporally distributed data, deeming it crucial to design training algorithms that enable efficient continual learning over decentralized private data. Decentralized learning allows serverless training with spatially distributed data. A fundamental barrier in such distributed learning is the high bandwidth cost of communicating model updates between agents. Moreover, existing works under this training paradigm are not inherently suitable for learning a temporal sequence of tasks while retaining the previously acquired knowledge. In this work, we propose CoDeC, a novel communication-efficient decentralized continual learning algorithm which addresses these challenges. We mitigate catastrophic forgetting while learning a task sequence in a decentralized learning setup by combining orthogonal gradient projection with gossip averaging across decentralized agents. Further, CoDeC includes a novel lossless communication compression scheme based on the gradient subspaces. We express layer-wise gradients as a linear combination of the basis vectors of these gradient subspaces and communicate the associated coefficients. We theoretically analyze the convergence rate for our algorithm and demonstrate through an extensive set of experiments that CoDeC successfully learns distributed continual tasks with minimal forgetting. The proposed compression scheme results in up to 4.8x reduction in communication costs with iso-performance as the full communication baseline.
Transformers have achieved great success in a wide variety of natural language processing (NLP) tasks due to the attention mechanism, which assigns an importance score for every word relative to other words in a sequence. However, these models are very large, often reaching hundreds of billions of parameters, and therefore require a large number of DRAM accesses. Hence, traditional deep neural network (DNN) accelerators such as GPUs and TPUs face limitations in processing Transformers efficiently. In-memory accelerators based on non-volatile memory promise to be an effective solution to this challenge, since they provide high storage density while performing massively parallel matrix vector multiplications within memory arrays. However, attention score computations, which are frequently used in Transformers (unlike CNNs and RNNs), require matrix vector multiplications (MVM) where both operands change dynamically for each input. As a result, conventional NVM-based accelerators incur high write latency and write energy when used for Transformers, and further suffer from the low endurance of most NVM technologies. To address these challenges, we present X-Former, a hybrid in-memory hardware accelerator that consists of both NVM and CMOS processing elements to execute transformer workloads efficiently. To improve the hardware utilization of X-Former, we also propose a sequence blocking dataflow, which overlaps the computations of the two processing elements and reduces execution time. Across several benchmarks, we show that X-Former achieves upto 85x and 7.5x improvements in latency and energy over a NVIDIA GeForce GTX 1060 GPU and upto 10.7x and 4.6x improvements in latency and energy over a state-of-the-art in-memory NVM accelerator.
In neural networks, continual learning results in gradient interference among sequential tasks, leading to catastrophic forgetting of old tasks while learning new ones. This issue is addressed in recent methods by storing the important gradient spaces for old tasks and updating the model orthogonally during new tasks. However, such restrictive orthogonal gradient updates hamper the learning capability of the new tasks resulting in sub-optimal performance. To improve new learning while minimizing forgetting, in this paper we propose a Scaled Gradient Projection (SGP) method, where we combine the orthogonal gradient projections with scaled gradient steps along the important gradient spaces for the past tasks. The degree of gradient scaling along these spaces depends on the importance of the bases spanning them. We propose an efficient method for computing and accumulating importance of these bases using the singular value decomposition of the input representations for each task. We conduct extensive experiments ranging from continual image classification to reinforcement learning tasks and report better performance with less training overhead than the state-of-the-art approaches.
Standard frame-based algorithms fail to retrieve accurate segmentation maps in challenging real-time applications like autonomous navigation, owing to the limited dynamic range and motion blur prevalent in traditional cameras. Event cameras address these limitations by asynchronously detecting changes in per-pixel intensity to generate event streams with high temporal resolution, high dynamic range, and no motion blur. However, event camera outputs cannot be directly used to generate reliable segmentation maps as they only capture information at the pixels in motion. To augment the missing contextual information, we postulate that fusing spatially dense frames with temporally dense events can generate semantic maps with fine-grained predictions. To this end, we propose HALSIE, a hybrid approach to learning segmentation by simultaneously leveraging image and event modalities. To enable efficient learning across modalities, our proposed hybrid framework comprises two input branches, a Spiking Neural Network (SNN) branch and a standard Artificial Neural Network (ANN) branch to process event and frame data respectively, while exploiting their corresponding neural dynamics. Our hybrid network outperforms the state-of-the-art semantic segmentation benchmarks on DDD17 and MVSEC datasets and shows comparable performance on the DSEC-Semantic dataset with upto 33.23$\times$ reduction in network parameters. Further, our method shows upto 18.92$\times$ improvement in inference cost compared to existing SOTA approaches, making it suitable for resource-constrained edge applications.
Spiking Neural Networks (SNNs) are bio-plausible models that hold great potential for realizing energy-efficient implementations of sequential tasks on resource-constrained edge devices. However, commercial edge platforms based on standard GPUs are not optimized to deploy SNNs, resulting in high energy and latency. While analog In-Memory Computing (IMC) platforms can serve as energy-efficient inference engines, they are accursed by the immense energy, latency, and area requirements of high-precision ADCs (HP-ADC), overshadowing the benefits of in-memory computations. We propose a hardware/software co-design methodology to deploy SNNs into an ADC-Less IMC architecture using sense-amplifiers as 1-bit ADCs replacing conventional HP-ADCs and alleviating the above issues. Our proposed framework incurs minimal accuracy degradation by performing hardware-aware training and is able to scale beyond simple image classification tasks to more complex sequential regression tasks. Experiments on complex tasks of optical flow estimation and gesture recognition show that progressively increasing the hardware awareness during SNN training allows the model to adapt and learn the errors due to the non-idealities associated with ADC-Less IMC. Also, the proposed ADC-Less IMC offers significant energy and latency improvements, $2-7\times$ and $8.9-24.6\times$, respectively, depending on the SNN model and the workload, compared to HP-ADC IMC.
Ransomware has appeared as one of the major global threats in recent days The alarming increasing rate of ransomware attacks and new ransomware variants intrigue the researchers to constantly examine the distinguishing traits of ransomware and refine their detection strategies Application Programming Interface API is a way for one program to collaborate with another API calls are the medium by which they communicate Ransomware uses this strategy to interact with the OS and makes a significantly higher number of calls in different sequences to ask for taking action This research work utilizes the frequencies of different API calls to detect and classify ransomware families First a WebCrawler is developed to automate collecting the Windows Portable Executable PE files of 15 different ransomware families By extracting different frequencies of 68 API calls we develop our dataset in the first phase of the two phase feature engineering process After selecting the most significant features in the second phase of the feature engineering process we deploy six Supervised Machine Learning models Naive Bayes Logistic Regression Random Forest Stochastic Gradient Descent K Nearest Neighbor and Support Vector Machine Then the performances of all the classifiers are compared to select the best model The results reveal that Logistic Regression can efficiently classify ransomware into their corresponding families securing 9915 accuracy Finally instead of relying on the Black box characteristic of the Machine Learning models we present the interpretability of our best performing model using SHAP values to ascertain the transparency and trustworthiness of the models prediction
Machine Learning (ML) has achieved enormous success in solving a variety of problems in computer vision, speech recognition, object detection, to name a few. The principal reason for this success is the availability of huge datasets for training deep neural networks (DNNs). However, datasets cannot be publicly released if they contain sensitive information such as medical records, and data privacy becomes a major concern. Encryption methods could be a possible solution, however their deployment on ML applications seriously impacts classification accuracy and results in substantial computational overhead. Alternatively, obfuscation techniques could be used, but maintaining a good trade-off between visual privacy and accuracy is challenging. In this paper, we propose a method to generate secure synthetic datasets from the original private datasets. Given a network with Batch Normalization (BN) layers pretrained on the original dataset, we first record the class-wise BN layer statistics. Next, we generate the synthetic dataset by optimizing random noise such that the synthetic data match the layer-wise statistical distribution of original images. We evaluate our method on image classification datasets (CIFAR10, ImageNet) and show that synthetic data can be used in place of the original CIFAR10/ImageNet data for training networks from scratch, producing comparable classification performance. Further, to analyze visual privacy provided by our method, we use Image Quality Metrics and show high degree of visual dissimilarity between the original and synthetic images. Moreover, we show that our proposed method preserves data-privacy under various privacy-leakage attacks including Gradient Matching Attack, Model Memorization Attack, and GAN-based Attack.
Domain-specific language understanding requires integrating multiple pieces of relevant contextual information. For example, we see both suicide and depression-related behavior (multiple contexts) in the text ``I have a gun and feel pretty bad about my life, and it wouldn't be the worst thing if I didn't wake up tomorrow''. Domain specificity in self-attention architectures is handled by fine-tuning on excerpts from relevant domain specific resources (datasets and external knowledge - medical textbook chapters on mental health diagnosis related to suicide and depression). We propose a modified self-attention architecture Knowledge-infused Self Attention Transformer (KSAT) that achieves the integration of multiple domain-specific contexts through the use of external knowledge sources. KSAT introduces knowledge-guided biases in dedicated self-attention layers for each knowledge source to accomplish this. In addition, KSAT provides mechanics for controlling the trade-off between learning from data and learning from knowledge. Our quantitative and qualitative evaluations show that (1) the KSAT architecture provides novel human-understandable ways to precisely measure and visualize the contributions of the infused domain contexts, and (2) KSAT performs competitively with other knowledge-infused baselines and significantly outperforms baselines that use fine-tuning for domain-specific tasks.
Prior works on event-based optical flow estimation have investigated several gradient-based learning methods to train neural networks for predicting optical flow. However, they do not utilize the fast data rate of event data streams and rely on a spatio-temporal representation constructed from a collection of events over a fixed period of time (often between two grayscale frames). As a result, optical flow is only evaluated at a frequency much lower than the rate data is produced by an event-based camera, leading to a temporally sparse optical flow estimation. To predict temporally dense optical flow, we cast the problem as a sequential learning task and propose a training methodology to train sequential networks for continuous prediction on an event stream. We propose two types of networks: one focused on performance and another focused on compute efficiency. We first train long-short term memory networks (LSTMs) on the DSEC dataset and demonstrated 10x temporally dense optical flow estimation over existing flow estimation approaches. The additional benefit of having a memory to draw long temporal correlations back in time results in a 19.7% improvement in flow prediction accuracy of LSTMs over similar networks with no memory elements. We subsequently show that the inherent recurrence of spiking neural networks (SNNs) enables them to learn and estimate temporally dense optical flow with 31.8% lesser parameters than LSTM, but with a slightly increased error. This demonstrates potential for energy-efficient implementation of fast optical flow prediction using SNNs.