We introduce a new method for internal replay that modulates the frequency of rehearsal based on the depth of the network. While replay strategies mitigate the effects of catastrophic forgetting in neural networks, recent works on generative replay show that performing the rehearsal only on the deeper layers of the network improves the performance in continual learning. However, the generative approach introduces additional computational overhead, limiting its applications. Motivated by the observation that earlier layers of neural networks forget less abruptly, we propose to update network layers with varying frequency using intermediate-level features during replay. This reduces the computational burden by omitting computations for both deeper layers of the generator and earlier layers of the main model. We name our method Progressive Latent Replay and show that it outperforms Internal Replay while using significantly fewer resources.
Continual Learning (CL) on time series data represents a promising but under-studied avenue for real-world applications. We propose two new CL benchmarks for Human State Monitoring. We carefully designed the benchmarks to mirror real-world environments in which new subjects are continuously added. We conducted an empirical evaluation to assess the ability of popular CL strategies to mitigate forgetting in our benchmarks. Our results show that, possibly due to the domain-incremental properties of our benchmarks, forgetting can be easily tackled even with a simple finetuning and that existing strategies struggle in accumulating knowledge over a fixed, held-out, test subject.
With increasing applications in areas such as biomedical information extraction pipelines and social media analytics, Named Entity Recognition (NER) has become an indispensable tool for knowledge extraction. However, with the gradual shift in language structure and vocabulary, NERs are plagued with distribution shifts, making them redundant or not as profitable without re-training. Re-training NERs based on Large Language Models (LLMs) from scratch over newly acquired data poses economic disadvantages. In contrast, re-training only with newly acquired data will result in Catastrophic Forgetting of previously acquired knowledge. Therefore, we propose NERDA-Con, a pipeline for training NERs with LLM bases by incorporating the concept of Elastic Weight Consolidation (EWC) into the NER fine-tuning NERDA pipeline. As we believe our work has implications to be utilized in the pipeline of continual learning and NER, we open-source our code as well as provide the fine-tuning library of the same name NERDA-Con at https://github.com/SupritiVijay/NERDA-Con and https://pypi.org/project/NERDA-Con/.
Synaptic memory consolidation has been heralded as one of the key mechanisms for supporting continual learning in neuromorphic Artificial Intelligence (AI) systems. Here we report that a Fowler-Nordheim (FN) quantum-tunneling device can implement synaptic memory consolidation similar to what can be achieved by algorithmic consolidation models like the cascade and the elastic weight consolidation (EWC) models. The proposed FN-synapse not only stores the synaptic weight but also stores the synapse's historical usage statistic on the device itself. We also show that the operation of the FN-synapse is near-optimal in terms of the synaptic lifetime and we demonstrate that a network comprising FN-synapses outperforms a comparable EWC network for a small benchmark continual learning task. With an energy footprint of femtojoules per synaptic update, we believe that the proposed FN-synapse provides an ultra-energy-efficient approach for implementing both synaptic memory consolidation and persistent learning.
Commercial and industrial deployments of robot fleets at Amazon, Nimble, Plus One, Waymo, and Zoox query remote human teleoperators when robots are at risk or unable to make task progress. With continual learning, interventions from the remote pool of humans can also be used to improve the robot fleet control policy over time. A central question is how to effectively allocate limited human attention. Prior work addresses this in the single-robot, single-human setting; we formalize the Interactive Fleet Learning (IFL) setting, in which multiple robots interactively query and learn from multiple human supervisors. We propose Return on Human Effort (ROHE) as a new metric and Fleet-DAgger, a family of IFL algorithms. We present an open-source IFL benchmark suite of GPU-accelerated Isaac Gym environments for standardized evaluation and development of IFL algorithms. We compare a novel Fleet-DAgger algorithm to 4 baselines with 100 robots in simulation. We also perform a physical block-pushing experiment with 4 ABB YuMi robot arms and 2 remote humans. Experiments suggest that the allocation of humans to robots significantly affects the performance of the fleet, and that the novel Fleet-DAgger algorithm can achieve up to 8.8x higher ROHE than baselines. See https://tinyurl.com/fleet-dagger for supplemental material.
In many real-world scenarios, data to train machine learning models become available over time. However, neural network models struggle to continually learn new concepts without forgetting what has been learnt in the past. This phenomenon is known as catastrophic forgetting and it is often difficult to prevent due to practical constraints, such as the amount of data that can be stored or the limited computation sources that can be used. Moreover, training large neural networks, such as Transformers, from scratch is very costly and requires a vast amount of training data, which might not be available in the application domain of interest. A recent trend indicates that dynamic architectures based on an expansion of the parameters can reduce catastrophic forgetting efficiently in continual learning, but this needs complex tuning to balance the growing number of parameters and barely share any information across tasks. As a result, they struggle to scale to a large number of tasks without significant overhead. In this paper, we validate in the computer vision domain a recent solution called Adaptive Distillation of Adapters (ADA), which is developed to perform continual learning using pre-trained Transformers and Adapters on text classification tasks. We empirically demonstrate on different classification tasks that this method maintains a good predictive performance without retraining the model or increasing the number of model parameters over the time. Besides it is significantly faster at inference time compared to the state-of-the-art methods.
While deep neural networks (DNNs) have achieved impressive classification performance in closed-world learning scenarios, they typically fail to generalize to unseen categories in dynamic open-world environments, in which the number of concepts is unbounded. In contrast, human and animal learners have the ability to incrementally update their knowledge by recognizing and adapting to novel observations. In particular, humans characterize concepts via exclusive (unique) sets of essential features, which are used for both recognizing known classes and identifying novelty. Inspired by natural learners, we introduce a Sparse High-level-Exclusive, Low-level-Shared feature representation (SHELS) that simultaneously encourages learning exclusive sets of high-level features and essential, shared low-level features. The exclusivity of the high-level features enables the DNN to automatically detect out-of-distribution (OOD) data, while the efficient use of capacity via sparse low-level features permits accommodating new knowledge. The resulting approach uses OOD detection to perform class-incremental continual learning without known class boundaries. We show that using SHELS for novelty detection results in statistically significant improvements over state-of-the-art OOD detection approaches over a variety of benchmark datasets. Further, we demonstrate that the SHELS model mitigates catastrophic forgetting in a class-incremental learning setting,enabling a combined novelty detection and accommodation framework that supports learning in open-world settings
Machine learning recently proved efficient in learning differential equations and dynamical systems from data. However, the data is commonly assumed to originate from a single never-changing system. In contrast, when modeling real-world dynamical processes, the data distribution often shifts due to changes in the underlying system dynamics. Continual learning of these processes aims to rapidly adapt to abrupt system changes without forgetting previous dynamical regimes. This work proposes an approach to continual learning based on reservoir computing, a state-of-the-art method for training recurrent neural networks on complex spatiotemporal dynamical systems. Reservoir computing fixes the recurrent network weights - hence these cannot be forgotten - and only updates linear projection heads to the output. We propose to train multiple competitive prediction heads concurrently. Inspired by neuroscience's predictive coding, only the most predictive heads activate, laterally inhibiting and thus protecting the inactive heads from forgetting induced by interfering parameter updates. We show that this multi-head reservoir minimizes interference and catastrophic forgetting on several dynamical systems, including the Van-der-Pol oscillator, the chaotic Lorenz attractor, and the high-dimensional Lorenz-96 weather model. Our results suggest that reservoir computing is a promising candidate framework for the continual learning of dynamical systems. We provide our code for data generation, method, and comparisons at \url{https://github.com/leonardbereska/multiheadreservoir}.
Online Continual learning is a challenging learning scenario where the model must learn from a non-stationary stream of data where each sample is seen only once. The main challenge is to incrementally learn while avoiding catastrophic forgetting, namely the problem of forgetting previously acquired knowledge while learning from new data. A popular solution in these scenario is to use a small memory to retain old data and rehearse them over time. Unfortunately, due to the limited memory size, the quality of the memory will deteriorate over time. In this paper we propose OLCGM, a novel replay-based continual learning strategy that uses knowledge condensation techniques to continuously compress the memory and achieve a better use of its limited size. The sample condensation step compresses old samples, instead of removing them like other replay strategies. As a result, the experiments show that, whenever the memory budget is limited compared to the complexity of the data, OLCGM improves the final accuracy compared to state-of-the-art replay strategies.
The goal of continual learning (CL) is to learn different tasks over time. The main desiderata associated with CL are to maintain performance on older tasks, leverage the latter to improve learning of future tasks, and to introduce minimal overhead in the training process (for instance, to not require a growing model or retraining). We propose the Neuro-Inspired Stability-Plasticity Adaptation (NISPA) architecture that addresses these desiderata through a sparse neural network with fixed density. NISPA forms stable paths to preserve learned knowledge from older tasks. Also, NISPA uses connection rewiring to create new plastic paths that reuse existing knowledge on novel tasks. Our extensive evaluation on EMNIST, FashionMNIST, CIFAR10, and CIFAR100 datasets shows that NISPA significantly outperforms representative state-of-the-art continual learning baselines, and it uses up to ten times fewer learnable parameters compared to baselines. We also make the case that sparsity is an essential ingredient for continual learning. The NISPA code is available at https://github.com/BurakGurbuz97/NISPA.
Current state-of-the-art vision-and-language models are evaluated on tasks either individually or in a multi-task setting, overlooking the challenges of continually learning (CL) tasks as they arrive. Existing CL benchmarks have facilitated research on task adaptation and mitigating "catastrophic forgetting", but are limited to vision-only and language-only tasks. We present CLiMB, a benchmark to study the challenge of learning multimodal tasks in a CL setting, and to systematically evaluate how upstream continual learning can rapidly generalize to new multimodal and unimodal tasks. CLiMB includes implementations of several CL algorithms and a modified Vision-Language Transformer (ViLT) model that can be deployed on both multimodal and unimodal tasks. We find that common CL methods can help mitigate forgetting during multimodal task learning, but do not enable cross-task knowledge transfer. We envision that CLiMB will facilitate research on a new class of CL algorithms for this challenging multimodal setting.
In this report we consider the following problem: Given a trained model that is partially faulty, can we correct its behaviour without having to train the model from scratch? In other words, can we ``debug" neural networks similar to how we address bugs in our mathematical models and standard computer code. We base our approach on the hypothesis that debugging can be treated as a two-task continual learning problem. In particular, we employ a modified version of a continual learning algorithm called Orthogonal Gradient Descent (OGD) to demonstrate, via two simple experiments on the MNIST dataset, that we can in-fact \textit{unlearn} the undesirable behaviour while retaining the general performance of the model, and we can additionally \textit{relearn} the appropriate behaviour, both without having to train the model from scratch.
Class incremental learning (CIL) algorithms aim to continually learn new object classes from incrementally arriving data while not forgetting past learned classes. The common evaluation protocol for CIL algorithms is to measure the average test accuracy across all classes learned so far -- however, we argue that solely focusing on maximizing the test accuracy may not necessarily lead to developing a CIL algorithm that also continually learns and updates the representations, which may be transferred to the downstream tasks. To that end, we experimentally analyze neural network models trained by CIL algorithms using various evaluation protocols in representation learning and propose new analysis methods. Our experiments show that most state-of-the-art algorithms prioritize high stability and do not significantly change the learned representation, and sometimes even learn a representation of lower quality than a naive baseline. However, we observe that these algorithms can still achieve high test accuracy because they enable a model to learn a classifier that closely resembles an estimated linear classifier trained for linear probing. Furthermore, the base model learned in the first task, which involves single-task learning, exhibits varying levels of representation quality across different algorithms, and this variance impacts the final performance of CIL algorithms. Therefore, we suggest that the representation-level evaluation should be considered as an additional recipe for more diverse evaluation for CIL algorithms.
We introduce a new training paradigm that enforces interval constraints on neural network parameter space to control forgetting. Contemporary Continual Learning (CL) methods focus on training neural networks efficiently from a stream of data, while reducing the negative impact of catastrophic forgetting, yet they do not provide any firm guarantees that network performance will not deteriorate uncontrollably over time. In this work, we show how to put bounds on forgetting by reformulating continual learning of a model as a continual contraction of its parameter space. To that end, we propose Hyperrectangle Training, a new training methodology where each task is represented by a hyperrectangle in the parameter space, fully contained in the hyperrectangles of the previous tasks. This formulation reduces the NP-hard CL problem back to polynomial time while providing full resilience against forgetting. We validate our claim by developing InterContiNet (Interval Continual Learning) algorithm which leverages interval arithmetic to effectively model parameter regions as hyperrectangles. Through experimental results, we show that our approach performs well in a continual learning setup without storing data from previous tasks.
Online few-shot learning describes a setting where models are trained and evaluated on a stream of data while learning emerging classes. While prior work in this setting has achieved very promising performance on instance classification when learning from data-streams composed of a single indoor environment, we propose to extend this setting to consider object classification on a series of several indoor environments, which is likely to occur in applications such as robotics. Importantly, our setting, which we refer to as online few-shot continual learning, injects the well-studied issue of catastrophic forgetting into the few-shot online learning paradigm. In this work, we benchmark several existing methods and adapted baselines within our setting, and show there exists a trade-off between catastrophic forgetting and online performance. Our findings motivate the need for future work in this setting, which can achieve better online performance without catastrophic forgetting.
Class-incremental learning (CIL) suffers from the notorious dilemma between learning newly added classes and preserving previously learned class knowledge. That catastrophic forgetting issue could be mitigated by storing historical data for replay, which yet would cause memory overheads as well as imbalanced prediction updates. To address this dilemma, we propose to leverage "free" external unlabeled data querying in continual learning. We first present a CIL with Queried Unlabeled Data (CIL-QUD) scheme, where we only store a handful of past training samples as anchors and use them to query relevant unlabeled examples each time. Along with new and past stored data, the queried unlabeled are effectively utilized, through learning-without-forgetting (LwF) regularizers and class-balance training. Besides preserving model generalization over past and current tasks, we next study the problem of adversarial robustness for CIL-QUD. Inspired by the recent success of learning robust models with unlabeled data, we explore a new robustness-aware CIL setting, where the learned adversarial robustness has to resist forgetting and be transferred as new tasks come in continually. While existing options easily fail, we show queried unlabeled data can continue to benefit, and seamlessly extend CIL-QUD into its robustified versions, RCIL-QUD. Extensive experiments demonstrate that CIL-QUD achieves substantial accuracy gains on CIFAR-10 and CIFAR-100, compared to previous state-of-the-art CIL approaches. Moreover, RCIL-QUD establishes the first strong milestone for robustness-aware CIL. Codes are available in https://github.com/VITA-Group/CIL-QUD.
In clinical practice, a segmentation network is often required to continually learn on a sequential data stream from multiple sites rather than a consolidated set, due to the storage cost and privacy restriction. However, during the continual learning process, existing methods are usually restricted in either network memorizability on previous sites or generalizability on unseen sites. This paper aims to tackle the challenging problem of Synchronous Memorizability and Generalizability (SMG) and to simultaneously improve performance on both previous and unseen sites, with a novel proposed SMG-learning framework. First, we propose a Synchronous Gradient Alignment (SGA) objective, which not only promotes the network memorizability by enforcing coordinated optimization for a small exemplar set from previous sites (called replay buffer), but also enhances the generalizability by facilitating site-invariance under simulated domain shift. Second, to simplify the optimization of SGA objective, we design a Dual-Meta algorithm that approximates the SGA objective as dual meta-objectives for optimization without expensive computation overhead. Third, for efficient rehearsal, we configure the replay buffer comprehensively considering additional inter-site diversity to reduce redundancy. Experiments on prostate MRI data sequentially acquired from six institutes demonstrate that our method can simultaneously achieve higher memorizability and generalizability over state-of-the-art methods. Code is available at https://github.com/jingyzhang/SMG-Learning.
In computational reinforcement learning, a growing body of work seeks to construct an agent's perception of the world through predictions of future sensations; predictions about environment observations are used as additional input features to enable better goal-directed decision-making. An open challenge in this line of work is determining from the infinitely many predictions that the agent could possibly make which predictions might best support decision-making. This challenge is especially apparent in continual learning problems where a single stream of experience is available to a singular agent. As a primary contribution, we introduce a meta-gradient descent process by which an agent learns 1) what predictions to make, 2) the estimates for its chosen predictions, and 3) how to use those estimates to generate policies that maximize future reward -- all during a single ongoing process of continual learning. In this manuscript we consider predictions expressed as General Value Functions: temporally extended estimates of the accumulation of a future signal. We demonstrate that through interaction with the environment an agent can independently select predictions that resolve partial-observability, resulting in performance similar to expertly specified GVFs. By learning, rather than manually specifying these predictions, we enable the agent to identify useful predictions in a self-supervised manner, taking a step towards truly autonomous systems.
Predictive machine learning models nowadays are often updated in a stateless and expensive way. The two main future trends for companies that want to build machine learning-based applications and systems are real-time inference and continual updating. Unfortunately, both trends require a mature infrastructure that is hard and costly to realize on-premise. This paper defines a novel software service and model delivery infrastructure termed Continual Learning-as-a-Service (CLaaS) to address these issues. Specifically, it embraces continual machine learning and continuous integration techniques. It provides support for model updating and validation tools for data scientists without an on-premise solution and in an efficient, stateful and easy-to-use manner. Finally, this CL model service is easy to encapsulate in any machine learning infrastructure or cloud system. This paper presents the design and implementation of a CLaaS instantiation, called LiquidBrain, evaluated in two real-world scenarios. The former is a robotic object recognition setting using the CORe50 dataset while the latter is a named category and attribute prediction using the DeepFashion-C dataset in the fashion domain. Our preliminary results suggest the usability and efficiency of the Continual Learning model services and the effectiveness of the solution in addressing real-world use-cases regardless of where the computation happens in the continuum Edge-Cloud.
The options framework in Hierarchical Reinforcement Learning breaks down overall goals into a combination of options or simpler tasks and associated policies, allowing for abstraction in the action space. Ideally, these options can be reused across different higher-level goals; indeed, such reuse is necessary to realize the vision of a continual learning agent that can effectively leverage its prior experience. Previous approaches have only proposed limited forms of transfer of prelearned options to new task settings. We propose a novel option indexing approach to hierarchical learning (OI-HRL), where we learn an affinity function between options and the items present in the environment. This allows us to effectively reuse a large library of pretrained options, in zero-shot generalization at test time, by restricting goal-directed learning to only those options relevant to the task at hand. We develop a meta-training loop that learns the representations of options and environments over a series of HRL problems, by incorporating feedback about the relevance of retrieved options to the higher-level goal. We evaluate OI-HRL in two simulated settings - the CraftWorld and AI2THOR environments - and show that we achieve performance competitive with oracular baselines, and substantial gains over a baseline that has the entire option pool available for learning the hierarchical policy.