Lukas Hedegaard

Lukas Hedegaard, Aman Alok, Juby Jose et al.

Adapters are a parameter-efficient alternative to fine-tuning, which augment a frozen base network to learn new tasks. Yet, the inference of the adapted model is often slower than the corresponding fine-tuned model. To improve on this, we propose Structured Pruning Adapters (SPAs), a family of compressing, task-switching network adapters, that accelerate and specialize networks using tiny parameter sets and structured pruning. Specifically, we propose a channel-based SPA and evaluate it with a suite of pruning methods on multiple computer vision benchmarks. Compared to regular structured pruning with fine-tuning, our channel-SPAs improve accuracy by 6.9% on average while using half the parameters at 90% pruned weights. Alternatively, they can learn adaptations with 17x fewer parameters at 70% pruning with 1.6% lower accuracy. Similarly, our block-SPA requires far fewer parameters than pruning with fine-tuning. Our experimental code and Python library of adapters are available at github.com/lukashedegaard/structured-pruning-adapters.

Lukas Hedegaard, Alexandros Iosifidis

We present Continual Inference, a Python library for implementing Continual Inference Networks (CINs) in PyTorch, a class of Neural Networks designed specifically for efficient inference in both online and batch processing scenarios. We offer a comprehensive introduction and guide to CINs and their implementation in practice, and provide best-practices and code examples for composing complex modules for modern Deep Learning. Continual Inference is readily downloadable via the Python Package Index and at \url{www.github.com/lukashedegaard/continual-inference}.

Lukas Hedegaard, Negar Heidari, Alexandros Iosifidis

Graph-based reasoning over skeleton data has emerged as a promising approach for human action recognition. However, the application of prior graph-based methods, which predominantly employ whole temporal sequences as their input, to the setting of online inference entails considerable computational redundancy. In this paper, we tackle this issue by reformulating the Spatio-Temporal Graph Convolutional Neural Network as a Continual Inference Network, which can perform step-by-step predictions in time without repeat frame processing. To evaluate our method, we create a continual version of ST-GCN, CoST-GCN, alongside two derived methods with different self-attention mechanisms, CoAGCN and CoS-TR. We investigate weight transfer strategies and architectural modifications for inference acceleration, and perform experiments on the NTU RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400 datasets. Retaining similar predictive accuracy, we observe up to 109x reduction in time complexity, on-hardware accelerations of 26x, and reductions in maximum allocated memory of 52% during online inference.

Lukas Hedegaard

The capabilities and adoption of deep neural networks (DNNs) grow at an exhilarating pace: Vision models accurately classify human actions in videos and identify cancerous tissue in medical scans as precisely than human experts; large language models answer wide-ranging questions, generate code, and write prose, becoming the topic of everyday dinner-table conversations. Even though their uses are exhilarating, the continually increasing model sizes and computational complexities have a dark side. The economic cost and negative environmental externalities of training and serving models is in evident disharmony with financial viability and climate action goals. Instead of pursuing yet another increase in predictive performance, this dissertation is dedicated to the improvement of neural network efficiency. Specifically, a core contribution addresses the efficiency aspects during online inference. Here, the concept of Continual Inference Networks (CINs) is proposed and explored across four publications. CINs extend prior state-of-the-art methods developed for offline processing of spatio-temporal data and reuse their pre-trained weights, improving their online processing efficiency by an order of magnitude. These advances are attained through a bottom-up computational reorganization and judicious architectural modifications. The benefit to online inference is demonstrated by reformulating several widely used network architectures into CINs, including 3D CNNs, ST-GCNs, and Transformer Encoders. An orthogonal contribution tackles the concurrent adaptation and computational acceleration of a large source model into multiple lightweight derived models. Drawing on fusible adapter networks and structured pruning, Structured Pruning Adapters achieve superior predictive accuracy under aggressive pruning using significantly fewer learned weights compared to fine-tuning with pruning.

Ginés Carreto Picón, Illia Oleksiienko, Lukas Hedegaard et al.

Transformers are widely used for their ability to capture data relations in sequence processing, with great success for a wide range of static tasks. However, the computational and memory footprint of their main component, i.e., the Scaled Dot-product Attention, is commonly overlooked. This makes their adoption in applications involving stream data processing with constraints in response latency, computational and memory resources infeasible. Some works have proposed methods to lower the computational cost of Transformers, i.e. low-rank approximations, sparsity in attention, and efficient formulations for Continual Inference. In this paper, we introduce a new formulation of the Scaled Dot-product Attention based on the Nyström approximation that is suitable for Continual Inference. In experiments on Online Audio Classification and Online Action Detection tasks, the proposed Continual Scaled Dot-product Attention can lower the number of operations by up to three orders of magnitude compared to the original Transformers while retaining the predictive performance of competing models.

Lukas Hedegaard, Arian Bakhtiarnia, Alexandros Iosifidis

Transformers in their common form are inherently limited to operate on whole token sequences rather than on one token at a time. Consequently, their use during online inference on time-series data entails considerable redundancy due to the overlap in successive token sequences. In this work, we propose novel formulations of the Scaled Dot-Product Attention, which enable Transformers to perform efficient online token-by-token inference on a continual input stream. Importantly, our modifications are purely to the order of computations, while the outputs and learned weights are identical to those of the original Transformer Encoder. We validate our Continual Transformer Encoder with experiments on the THUMOS14, TVSeries and GTZAN datasets with remarkable results: Our Continual one- and two-block architectures reduce the floating point operations per prediction by up to 63x and 2.6x, respectively, while retaining predictive performance.

Lukas Hedegaard, Alexandros Iosifidis

We introduce Continual 3D Convolutional Neural Networks (Co3D CNNs), a new computational formulation of spatio-temporal 3D CNNs, in which videos are processed frame-by-frame rather than by clip. In online tasks demanding frame-wise predictions, Co3D CNNs dispense with the computational redundancies of regular 3D CNNs, namely the repeated convolutions over frames, which appear in overlapping clips. We show that Continual 3D CNNs can reuse preexisting 3D-CNN weights to reduce the per-prediction floating point operations (FLOPs) in proportion to the temporal receptive field while retaining similar memory requirements and accuracy. This is validated with multiple models on Kinetics-400 and Charades with remarkable results: CoX3D models attain state-of-the-art complexity/accuracy trade-offs on Kinetics-400 with 12.1-15.3x reductions of FLOPs and 2.3-3.8% improvements in accuracy compared to regular X3D models while reducing peak memory consumption by up to 48%. Moreover, we investigate the transient response of Co3D CNNs at start-up and perform extensive benchmarks of on-hardware processing characteristics for publicly available 3D CNNs.

Lukas Hedegaard, Omar Ali Sheikh-Omar, Alexandros Iosifidis

Domain Adaptation is the process of alleviating distribution gaps between data from different domains. In this paper, we show that Domain Adaptation methods using pair-wise relationships between source and target domain data can be formulated as a Graph Embedding in which the domain labels are incorporated into the structure of the intrinsic and penalty graphs. Specifically, we analyse the loss functions of three existing state-of-the-art Supervised Domain Adaptation methods and demonstrate that they perform Graph Embedding. Moreover, we highlight some generalisation and reproducibility issues related to the experimental setup commonly used to demonstrate the few-shot learning capabilities of these methods. To assess and compare Supervised Domain Adaptation methods accurately, we propose a rectified evaluation protocol, and report updated benchmarks on the standard datasets Office31 (Amazon, DSLR, and Webcam), Digits (MNIST, USPS, SVHN, and MNIST-M) and VisDA (Synthetic, Real).