Look At Me, No Replay! SurpriseNet: Anomaly Detection Inspired Class Incremental Learning

Continual learning aims to create artificial neural networks capable of accumulating knowledge and skills through incremental training on a sequence of tasks. The main challenge of continual learning is catastrophic interference, wherein new knowledge overrides or interferes with past knowledge, leading to forgetting. An associated issue is the problem of learning "cross-task knowledge," where models fail to acquire and retain knowledge that helps differentiate classes across task boundaries. A common solution to both problems is "replay," where a limited buffer of past instances is utilized to learn cross-task knowledge and mitigate catastrophic interference. However, a notable drawback of these methods is their tendency to overfit the limited replay buffer. In contrast, our proposed solution, SurpriseNet, addresses catastrophic interference by employing a parameter isolation method and learning cross-task knowledge using an auto-encoder inspired by anomaly detection. SurpriseNet is applicable to both structured and unstructured data, as it does not rely on image-specific inductive biases. We have conducted empirical experiments demonstrating the strengths of SurpriseNet on various traditional vision continual-learning benchmarks, as well as on structured data datasets. Source code made available at https://doi.org/10.5281/zenodo.8247906 and https://github.com/tachyonicClock/SurpriseNet-CIKM-23

A simple but strong baseline for online continual learning: Repeated Augmented Rehearsal

Online continual learning (OCL) aims to train neural networks incrementally from a non-stationary data stream with a single pass through data. Rehearsal-based methods attempt to approximate the observed input distributions over time with a small memory and revisit them later to avoid forgetting. Despite its strong empirical performance, rehearsal methods still suffer from a poor approximation of the loss landscape of past data with memory samples. This paper revisits the rehearsal dynamics in online settings. We provide theoretical insights on the inherent memory overfitting risk from the viewpoint of biased and dynamic empirical risk minimization, and examine the merits and limits of repeated rehearsal. Inspired by our analysis, a simple and intuitive baseline, Repeated Augmented Rehearsal (RAR), is designed to address the underfitting-overfitting dilemma of online rehearsal. Surprisingly, across four rather different OCL benchmarks, this simple baseline outperforms vanilla rehearsal by 9%-17% and also significantly improves state-of-the-art rehearsal-based methods MIR, ASER, and SCR. We also demonstrate that RAR successfully achieves an accurate approximation of the loss landscape of past data and high-loss ridge aversion in its learning trajectory. Extensive ablation studies are conducted to study the interplay between repeated and augmented rehearsal and reinforcement learning (RL) is applied to dynamically adjust the hyperparameters of RAR to balance the stability-plasticity trade-off online.

Cross-domain Few-shot Meta-learning Using Stacking

Cross-domain few-shot meta-learning (CDFSML) addresses learning problems where knowledge needs to be transferred from several source domains into an instance-scarce target domain with an explicitly different input distribution. Recently published CDFSML methods generally construct a "universal model" that combines knowledge of multiple source domains into one backbone feature extractor. This enables efficient inference but necessitates re-computation of the backbone whenever a new source domain is added. Moreover, state-of-the-art methods derive their universal model from a collection of backbones -- normally one for each source domain -- and the backbones may be constrained to have the same architecture as the universal model. We propose a CDFSML method that is inspired by the classic stacking approach to meta learning. It imposes no constraints on the backbones' architecture or feature shape and does not incur the computational overhead of (re-)computing a universal model. Given a target-domain task, it fine-tunes each backbone independently, uses cross-validation to extract meta training data from the task's instance-scarce support set, and learns a simple linear meta classifier from this data. We evaluate our stacking approach on the well-known Meta-Dataset benchmark, targeting image classification with convolutional neural networks, and show that it often yields substantially higher accuracy than competing methods.

Balancing Performance and Energy Consumption of Bagging Ensembles for the Classification of Data Streams in Edge Computing

In recent years, the Edge Computing (EC) paradigm has emerged as an enabling factor for developing technologies like the Internet of Things (IoT) and 5G networks, bridging the gap between Cloud Computing services and end-users, supporting low latency, mobility, and location awareness to delay-sensitive applications. Most solutions in EC employ machine learning (ML) methods to perform data classification and other information processing tasks on continuous and evolving data streams. Usually, such solutions have to cope with vast amounts of data that come as data streams while balancing energy consumption, latency, and the predictive performance of the algorithms. Ensemble methods achieve remarkable predictive performance when applied to evolving data streams due to the combination of several models and the possibility of selective resets. This work investigates strategies for optimizing the performance (i.e., delay, throughput) and energy consumption of bagging ensembles to classify data streams. The experimental evaluation involved six state-of-art ensemble algorithms (OzaBag, OzaBag Adaptive Size Hoeffding Tree, Online Bagging ADWIN, Leveraging Bagging, Adaptive RandomForest, and Streaming Random Patches) applying five widely used machine learning benchmark datasets with varied characteristics on three computer platforms. Such strategies can significantly reduce energy consumption in 96% of the experimental scenarios evaluated. Despite the trade-offs, it is possible to balance them to avoid significant loss in predictive performance.

Improving the performance of bagging ensembles for data streams through mini-batching

Often, machine learning applications have to cope with dynamic environments where data are collected in the form of continuous data streams with potentially infinite length and transient behavior. Compared to traditional (batch) data mining, stream processing algorithms have additional requirements regarding computational resources and adaptability to data evolution. They must process instances incrementally because the data's continuous flow prohibits storing data for multiple passes. Ensemble learning achieved remarkable predictive performance in this scenario. Implemented as a set of (several) individual classifiers, ensembles are naturally amendable for task parallelism. However, the incremental learning and dynamic data structures used to capture the concept drift increase the cache misses and hinder the benefit of parallelism. This paper proposes a mini-batching strategy that can improve memory access locality and performance of several ensemble algorithms for stream mining in multi-core environments. With the aid of a formal framework, we demonstrate that mini-batching can significantly decrease the reuse distance (and the number of cache misses). Experiments on six different state-of-the-art ensemble algorithms applying four benchmark datasets with varied characteristics show speedups of up to 5X on 8-core processors. These benefits come at the expense of a small reduction in predictive performance.

Improving Predictions of Tail-end Labels using Concatenated BioMed-Transformers for Long Medical Documents

Multi-label learning predicts a subset of labels from a given label set for an unseen instance while considering label correlations. A known challenge with multi-label classification is the long-tailed distribution of labels. Many studies focus on improving the overall predictions of the model and thus do not prioritise tail-end labels. Improving the tail-end label predictions in multi-label classifications of medical text enables the potential to understand patients better and improve care. The knowledge gained by one or more infrequent labels can impact the cause of medical decisions and treatment plans. This research presents variations of concatenated domain-specific language models, including multi-BioMed-Transformers, to achieve two primary goals. First, to improve F1 scores of infrequent labels across multi-label problems, especially with long-tail labels; second, to handle long medical text and multi-sourced electronic health records (EHRs), a challenging task for standard transformers designed to work on short input sequences. A vital contribution of this research is new state-of-the-art (SOTA) results obtained using TransformerXL for predicting medical codes. A variety of experiments are performed on the Medical Information Mart for Intensive Care (MIMIC-III) database. Results show that concatenated BioMed-Transformers outperform standard transformers in terms of overall micro and macro F1 scores and individual F1 scores of tail-end labels, while incurring lower training times than existing transformer-based solutions for long input sequences.

Predicting COVID-19 Patient Shielding: A Comprehensive Study

There are many ways machine learning and big data analytics are used in the fight against the COVID-19 pandemic, including predictions, risk management, diagnostics, and prevention. This study focuses on predicting COVID-19 patient shielding -- identifying and protecting patients who are clinically extremely vulnerable from coronavirus. This study focuses on techniques used for the multi-label classification of medical text. Using the information published by the United Kingdom NHS and the World Health Organisation, we present a novel approach to predicting COVID-19 patient shielding as a multi-label classification problem. We use publicly available, de-identified ICU medical text data for our experiments. The labels are derived from the published COVID-19 patient shielding data. We present an extensive comparison across 12 multi-label classifiers from the simple binary relevance to neural networks and the most recent transformers. To the best of our knowledge this is the first comprehensive study, where such a range of multi-label classifiers for medical text are considered. We highlight the benefits of various approaches, and argue that, for the task at hand, both predictive accuracy and processing time are essential.

Better Self-training for Image Classification through Self-supervision

Self-training is a simple semi-supervised learning approach: Unlabelled examples that attract high-confidence predictions are labelled with their predictions and added to the training set, with this process being repeated multiple times. Recently, self-supervision -- learning without manual supervision by solving an automatically-generated pretext task -- has gained prominence in deep learning. This paper investigates three different ways of incorporating self-supervision into self-training to improve accuracy in image classification: self-supervision as pretraining only, self-supervision performed exclusively in the first iteration of self-training, and self-supervision added to every iteration of self-training. Empirical results on the SVHN, CIFAR-10, and PlantVillage datasets, using both training from scratch, and Imagenet-pretrained weights, show that applying self-supervision only in the first iteration of self-training can greatly improve accuracy, for a modest increase in computation time.

Transfer of Pretrained Model Weights Substantially Improves Semi-Supervised Image Classification

Deep neural networks produce state-of-the-art results when trained on a large number of labeled examples but tend to overfit when small amounts of labeled examples are used for training. Creating a large number of labeled examples requires considerable resources, time, and effort. If labeling new data is not feasible, so-called semi-supervised learning can achieve better generalisation than purely supervised learning by employing unlabeled instances as well as labeled ones. The work presented in this paper is motivated by the observation that transfer learning provides the opportunity to potentially further improve performance by exploiting models pretrained on a similar domain. More specifically, we explore the use of transfer learning when performing semi-supervised learning using self-learning. The main contribution is an empirical evaluation of transfer learning using different combinations of similarity metric learning methods and label propagation algorithms in semi-supervised learning. We find that transfer learning always substantially improves the model's accuracy when few labeled examples are available, regardless of the type of loss used for training the neural network. This finding is obtained by performing extensive experiments on the SVHN, CIFAR10, and Plant Village image classification datasets and applying pretrained weights from Imagenet for transfer learning.

Semi-Supervised Learning using Siamese Networks

Neural networks have been successfully used as classification models yielding state-of-the-art results when trained on a large number of labeled samples. These models, however, are more difficult to train successfully for semi-supervised problems where small amounts of labeled instances are available along with a large number of unlabeled instances. This work explores a new training method for semi-supervised learning that is based on similarity function learning using a Siamese network to obtain a suitable embedding. The learned representations are discriminative in Euclidean space, and hence can be used for labeling unlabeled instances using a nearest-neighbor classifier. Confident predictions of unlabeled instances are used as true labels for retraining the Siamese network on the expanded training set. This process is applied iteratively. We perform an empirical study of this iterative self-training algorithm. For improving unlabeled predictions, local learning with global consistency [22] is also evaluated.