16 September 2021 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Pretrained deep models hold their learnt knowledge in the form of model parameters. These parameters act as “memory” for the trained models and help them generalize well on unseen data. However, in absence of training data, the utility of a trained model is merely limited to either inference or better initialization towards a target task. In this paper, we go further and extract synthetic data by leveraging the learnt model parameters. We dub them Data Impressions , which act as proxy to the training data and can be used to realize a variety of tasks. These are useful in scenarios where only the pretrained models are available and the training data is not shared (e.g., due to privacy or sensitivity concerns). We show the applicability of data impressions in solving several computer vision tasks such as unsupervised domain adaptation, continual learning as well as knowledge distillation. We also study the adversarial robustness of lightweight models trained via knowledge distillation using these data impressions. Further, we demonstrate the efficacy of data impressions in generating data-free Universal Adversarial Perturbations (UAPs) with better fooling rates. Extensive experiments performed on benchmark datasets demonstrate competitive performance achieved using data impressions in absence of original training data.
4 December 2019 in IEEE Transactions on Information Forensics and Security
4 July 2023 in International Journal of Computer Vision
29 July 2021 in Frontiers in Neuroscience
Event-basedcameras are bio-inspired novel sensors that asynchronously record changes in illumination in the form of events. This principle results in significant advantages over conventional cameras, such as low power utilization, high dynamic range, and no motion blur. Moreover, by design, such cameras encode only the relative motion between the scene and the sensor and not the static background to yield a very sparse data structure. In this paper, we leverage these advantages of an event camera toward a critical vision application—video anomaly detection. We propose an anomaly detection solution in the event domain with a conditional Generative Adversarial Network (cGAN) made up of sparse submanifold convolution layers. Video analytics tasks such as anomaly detection depend on the motion history at each pixel. To enable this, we also put forward a generic unsupervised deep learning solution to learn a novel memory surface known as Deep Learning (DL) memory surface. DL memory surface encodes the temporal information readily available from these sensors while retaining the sparsity of event data. Since there is no existing dataset for anomaly detection in the event domain, we also provide an anomaly detection event dataset with a set of anomalies. We empirically validate our anomaly detection architecture, composed of sparse convolutional layers, on this proposed and online dataset. Careful analysis of the anomaly detection network reveals that the presented method results in a massive reduction in computational complexity with good performance compared to previous state-of-the-art conventional frame-based anomaly detection networks.
13 January 2025 in IEEE Transactions on Cloud Computing
Containerization has become popular for the deployment of applications on public clouds. Large enterprises may host 100s of applications on 1000s containers that are placed onto Virtual Machines (VMs). Such placement decisions happen continuously as applications are updated by DevOps pipelines that deploy the containers. Managing the placement of container resource requests onto the available capacities of VMs needs to be cost-efficient. This is well-studied, and usually modelled as a multi-dimensional Vector Bin-packing Problem (VBP). Many heuristics, and recently machine learning approaches, have been developed to solve this NP-hard problem for real-time decisions. We propose CARL, a novel approach to solve VBP through Adversarial Reinforcement Learning (RL) for cost minimization. It mimics the placement behavior of an offline semi-optimal VBP solver (teacher), while automatically learning a reward function for reducing the VM costs which out-performs the teacher. It requires limited historical container workload traces to train, and is resilient to changes in the workload distribution during inferencing. We extensively evaluate CARL on workloads derived from realistic traces from Google and Alibaba for the placement of 5k–10k container requests onto 2k–8k VMs, and compare it with classic heuristics and state-of-the-art RL methods. (1) CARL is fast, e.g., making placement decisions at ≈ 1900 requests/sec onto 8, 900 candidate VMs. (2) It is efficient, achieving ≈ 16% lower VM costs than classic and contemporary RL methods. (3) It is robust to changes in the workload, offering competitive results even when the resource needs or inter-arrival time of the container requests skew from the training workload.