Slow wave sleep is associated with forgetting in patients with temporal lobe epilepsy.

While time spent in slow wave sleep (SWS) after learning promotes memory consolidation in the healthy brain, it is unclear if the same benefit is obtained in patients with temporal lobe epilepsy (TLE). Interictal epileptiform discharges (IEDs) are potentiated during SWS and thus may disrupt memory consolidation processes thought to depend on hippocampal-neocortical interactions. Here, we explored the relationship between SWS, IEDs, and overnight forgetting in patients with TLE. Nineteen patients with TLE studied object-scene pairs and memory was tested across a day of wakefulness (6 hrs) and across a night of sleep (16 hrs) while undergoing continuous scalp EEG monitoring. We found that time spent in SWS after learning was related to greater forgetting overnight. Longer duration in SWS and number of IEDs were each associated with greater forgetting, although the number of IEDs did not mediate the relationship between SWS and memory. Further research, particularly with intracranial recordings, is required to identify the mechanisms by which SWS and IEDs can be pathological to sleep-dependent memory consolidation in patients with TLE.

Towards Robust Decision-Making for Autonomous Highway Driving Based on Safe Reinforcement Learning.

Reinforcement Learning (RL) methods are regarded as effective for designing autonomous driving policies. However, even when RL policies are trained to convergence, ensuring their robust safety remains a challenge, particularly in long-tail data. Therefore, decision-making based on RL must adequately consider potential variations in data distribution. This paper presents a framework for highway autonomous driving decisions that prioritizes both safety and robustness. Utilizing the proposed Replay Buffer Constrained Policy Optimization (RECPO) method, this framework updates RL strategies to maximize rewards while ensuring that the policies always remain within safety constraints. We incorporate importance sampling techniques to collect and store data in a Replay buffer during agent operation, allowing the reutilization of data from old policies for training new policy models, thus mitigating potential catastrophic forgetting. Additionally, we transform the highway autonomous driving decision problem into a Constrained Markov Decision Process (CMDP) and apply our proposed RECPO for training, optimizing highway driving policies. Finally, we deploy our method in the CARLA simulation environment and compare its performance in typical highway scenarios against traditional CPO, current advanced strategies based on Deep Deterministic Policy Gradient (DDPG), and IDM + MOBIL (Intelligent Driver Model and the model for minimizing overall braking induced by lane changes). The results show that our framework significantly enhances model convergence speed, safety, and decision-making stability, achieving a zero-collision rate in highway autonomous driving.

Continual pre-training mitigates forgetting in language and vision.

Pre-trained models are commonly used in Continual Learning to initialize the model before training on the stream of non-stationary data. However, pre-training is rarely applied during Continual Learning. We investigate the characteristics of the Continual Pre-Training scenario, where a model is continually pre-trained on a stream of incoming data and only later fine-tuned to different downstream tasks. We introduce an evaluation protocol for Continual Pre-Training which monitors forgetting against a Forgetting Control dataset not present in the continual stream. We disentangle the impact on forgetting of 3 main factors: the input modality (NLP, Vision), the architecture type (Transformer, ResNet) and the pre-training protocol (supervised, self-supervised). Moreover, we propose a Sample-Efficient Pre-training method (SEP) that speeds up the pre-training phase. We show that the pre-training protocol is the most important factor accounting for forgetting. Surprisingly, we discovered that self-supervised continual pre-training in both NLP and Vision is sufficient to mitigate forgetting without the use of any Continual Learning strategy. Other factors, like model depth, input modality and architecture type are not as crucial.

Multicenter privacy-preserving model training for deep learning brain metastases autosegmentation.

OBJECTIVES: This work aims to explore the impact of multicenter data heterogeneity on deep learning brain metastases (BM) autosegmentation performance, and assess the efficacy of an incremental transfer learning technique, namely learning without forgetting (LWF), to improve model generalizability without sharing raw data. MATERIALS AND METHODS: A total of six BM datasets from University Hospital Erlangen (UKER), University Hospital Zurich (USZ), Stanford, UCSF, New York University (NYU), and BraTS Challenge 2023 were used. First, the performance of the DeepMedic network for BM autosegmentation was established for exclusive single-center training and mixed multicenter training, respectively. Subsequently privacy-preserving bilateral collaboration was evaluated, where a pretrained model is shared to another center for further training using transfer learning (TL) either with or without LWF. RESULTS: For single-center training, average F1 scores of BM detection range from 0.625 (NYU) to 0.876 (UKER) on respective single-center test data. Mixed multicenter training notably improves F1 scores at Stanford and NYU, with negligible improvement at other centers. When the UKER pretrained model is applied to USZ, LWF achieves a higher average F1 score (0.839) than naive TL (0.570) and single-center training (0.688) on combined UKER and USZ test data. Naive TL improves sensitivity and contouring accuracy, but compromises precision. Conversely, LWF demonstrates commendable sensitivity, precision and contouring accuracy. When applied to Stanford, similar performance was observed. CONCLUSION: Data heterogeneity (e.g., variations in metastases density, spatial distribution, and image spatial resolution across centers) results in varying performance in BM autosegmentation, posing challenges to model generalizability. LWF is a promising approach to peer-to-peer privacy-preserving model training.

Effect of levels-of-processing on rates of forgetting.

The levels-of-processing (LOP) framework, proposing that deep processing yields superior retention, has provided an important paradigm for memory research and a practical means of improving learning. However, the available levels-of-processing literature focuses on immediate memory performance. It is assumed within the LOP framework that deep processing will lead to slower forgetting than will shallow processing. However, it is unclear whether, or how, the initial level of processing affects the forgetting slopes over longer retention intervals. The present three experiments were designed to explore whether items encoded at qualitatively different LOP are forgotten at different rates. In the first two experiments, depth of processing was manipulated within-participants at encoding under deep and shallow conditions (semantic vs. rhyme judgement in Experiment 1; semantic vs. consonant-vowel pattern decision in Experiment 2). Recognition accuracy (d prime) was measured between-participants immediately after learning and at 30-min, 2-h, and 24-h delays. The third experiment employed a between-participants design, contrasting the rates of forgetting following semantic and phonological (rhyme) processing at immediate, 30-min, 2-h, and 6-h delays. Results from the three experiments consistently demonstrated a large effect size of levels of processing on immediate performance and a medium-to-large level effect size on delayed recognition, but crucially no LOP × delay group interaction. Analysis of the retention curves revealed no significant differences between the slopes of forgetting for deep and shallow processing. These results suggest that the rates of forgetting are independent of the qualitatively distinct encoding operations manipulated by levels of processing.

Less confidence, less forgetting: Learning with a humbler teacher in exemplar-free Class-Incremental learning.

Class-Incremental learning (CIL) is challenging due to catastrophic forgetting (CF), which escalates in exemplar-free scenarios. To mitigate CF, Knowledge Distillation (KD), which leverages old models as teacher models, has been widely employed in CIL. However, based on a case study, our investigation reveals that the teacher model exhibits over-confidence in unseen new samples. In this article, we conduct empirical experiments and provide theoretical analysis to investigate the over-confident phenomenon and the impact of KD in exemplar-free CIL, where access to old samples is unavailable. Building on our analysis, we propose a novel approach, Learning with Humbler Teacher, by systematically selecting an appropriate checkpoint model as a humbler teacher to mitigate CF. Furthermore, we explore utilizing the nuclear norm to obtain an appropriate temporal ensemble to enhance model stability. Notably, LwHT outperforms the state-of-the-art approach by a significant margin of 10.41%, 6.56%, and 4.31% in various settings while demonstrating superior model plasticity.

Influence of degree of learning on rate of forgetting of tonal sequences.

Initial performance is frequently equated in studies that compare forgetting rates across groups. However, since the encoding capacity of different groups can be different, some procedures to match initial degree of learning need to be implemented, adding confounding variables such as longer exposures to the material, which would create memories of a different age. Slamecka and McElree Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 384-397, (1983) and our previous work found that the rate of forgetting was independent from initial degree of learning using verbal material. The present study seeks to determine whether this pattern holds true when undertaken with nonverbal material. In two experiments, we manipulate initial degree of learning by varying the number of presentations of the material and studying the effect on the forgetting rates. A set of 30 tonal sequences were presented to young, healthy participants either once or three times. Forgetting was evaluated in a yes/no recognition paradigm immediately and 1 hour or 24 hours after the study phase. A different subset of 10 sequences was tested along with 10 nontargets at each retention interval. The results of these experiments showed that initial acquisition was modulated by the number of repetitions. However, the forgetting rates were independent of initial degree of learning. These results are in keeping with the pattern found by Slamecka and McElree, and in our own previous studies. They suggest that the pattern of parallel forgetting after different levels of initial learning is not limited to verbal material.

Energy-efficient dynamic sensor time series classification for edge health devices.

BACKGROUND AND OBJECTIVE: Time series data plays a crucial role in the realm of the Internet of Things Medical (IoMT). Through machine learning (ML) algorithms, online time series classification in IoMT systems enables reliable real-time disease detection. Deploying ML algorithms on edge health devices can reduce latency and safeguard patients' privacy. However, the limited computational resources of these devices underscore the need for more energy-efficient algorithms. Furthermore, online time series classification inevitably faces the challenges of concept drift (CD) and catastrophic forgetting (CF). To address these challenges, this study proposes an energy-efficient Online Time series classification algorithm that can solve CF and CD for health devices, called OTCD. METHODS: OTCD first detects the appearance of concept drift and performs prototype updates to mitigate its impact. Afterward, it standardizes the potential space distribution and selectively preserves key training parameters to address CF. This approach reduces the required memory and enhances energy efficiency. To evaluate the performance of the proposed model in real-time health monitoring tasks, we utilize electrocardiogram (ECG) and photoplethysmogram (PPG) data. By adopting various feature extractors, three arrhythmia classification models are compared. To assess the energy efficiency of OTCD, we conduct runtime tests on each dataset. Additionally, the OTCD is compared with state-of-the-art (SOTA) dynamic time series classification models for performance evaluation. RESULTS: The OTCD algorithm outperforms existing SOTA time series classification algorithms in IoMT. In particular, OTCD is on average 2.77% to 14.74% more accurate than other models on the MIT-BIH arrhythmia dataset. Additionally, it consumes low memory (1 KB) and performs computations at a rate of 0.004 GFLOPs per second, leading to energy savings and high time efficiency. CONCLUSION: Our proposed algorithm, OTCD, enables efficient real-time classification of medical time series on edge health devices. Experimental results demonstrate its significant competitiveness, offering promising prospects for safe and reliable healthcare.

Long-term forgetting is independent of age in healthy children and adolescents.

Introduction: In clinical neuropsychology, the phenomenon of accelerated long-term forgetting (ALF) has advanced to be a marker for subtle but clinically relevant memory problems associated with a range of neurological conditions. The normal developmental trajectory of long-term memory, in this case, memory recall after 1 week, and the influence of cognitive variables such as intelligence have not extensively been described, which is a drawback for the use of accelerated long-term forgetting measures in pediatric neuropsychology. Methods: In this clinical observation study, we analyzed the normal developmental trajectory of verbal memory recall after 1 week in healthy children and adolescents. We hypothesized that 1-week recall and 1-week forgetting would be age-dependent and correlate with other cognitive functions such as intelligence and working memory. Sixty-three healthy participants between the ages of 8 and 16 years completed a newly developed auditory verbal learning test (WoMBAT) and the WISC-V intelligence test (General Ability Index, GAI). Using these tests, 1 week recall and 1 week forgetting have been studied in relation to GAI, verbal learning performance, and verbal working memory. Results: Neither 1-week recall nor 1-week forgetting seems to be age-dependent. They are also not significantly predicted by other cognitive functions such as GAI or working memory. Instead, the ability to recall a previously memorized word list after 7 days seems to depend solely on the initial learning capacity. Conclusion: In the clinical setting, this finding can help interpret difficulties in free recall after 7 days or more since they can probably not be attributed to young age or low intelligence.

Impaired episodic verbal memory recall after 1 week and elevated forgetting in children after mild traumatic brain injury - results from a short-term longitudinal study.

Objective: There is preliminary evidence that children after traumatic brain injury (TBI) have accelerated long-term forgetting (ALF), i.e., an adequate learning and memory performance in standardized memory tests, but an excessive rate of forgetting over delays of days or weeks. The main aim of this study was to investigate episodic memory performance, including delayed retrieval 1 week after learning, in children after mild TBI (mTBI). Methods: This prospective study with two time-points (T1: 1 week after injury and T2: 3-6 months after injury), included data of 64 children after mTBI and 57 healthy control children aged between 8 and 16 years. We assessed episodic learning and memory using an auditory word learning test and compared executive functions (interference control, working memory, semantic fluency and flexibility) and divided attention between groups. We explored correlations between memory performance and executive functions. Furthermore, we examined predictive factors for delayed memory retrieval 1 week after learning as well as for forgetting over time. Results: Compared to healthy controls, patients showed an impaired delayed recall and recognition performance 3-6 months after injury. Executive functions, but not divided attention, were reduced in children after mTBI. Furthermore, parents rated episodic memory as impaired 3-6 months after injury. Additionally, verbal learning and group, but not executive functions, were predictive for delayed recall performance at both time-points, whereas forgetting was predicted by group. Discussion: Delayed recall and forgetting over time were significantly different between groups, both post-acutely and in the chronic phase after pediatric mTBI, even in a very mildly injured patient sample. Delayed memory performance should be included in clinical evaluations of episodic memory and further research is needed to understand the mechanisms of ALF.

Enhancing multi-scenario applicability of freeway variable speed limit control strategies using continual learning.

Variable speed limit (VSL) control benefits freeway operations through dynamic speed limit adjustment strategies for specific operation scenarios, such as traffic jams, secondary crash prevention, etc. To develop optimal strategies, deep reinforcement learning (DRL) has been employed to map the traffic operation status to speed limits with the corresponding control effects. Then, VSL control strategies were obtained based upon memories of these complex mapping relationships. However, under multi-scenario conditions, DRL trained VSL faces the challenge of performance decay, where the control strategy effects drop sharply for early trained "old scenarios". This so-called scenario forgetting problem is attributed to the fact that DRL would forget the learned old scenario mapping memories after new scenario trainings. To tackle this issue, a continual learning approach has been introduced in this study to enhance the multi-scenario applicability of VSL control strategies. Specifically, a gradient projection memory (GPM) based neural network parameter updating method was proposed to keep the mapping memories of old scenarios during new scenario trainings by imposing constraints on the direction of gradient updates for new tasks. The proposed method was evaluated using three typical freeway operation scenarios developed in the simulation platform SUMO. Experimental results showed that the continual learning approach has substantially reduced the performance decay in old scenarios by 17.76% (valued using backward transfer metrics). Furthermore, the multi-scenario VSL control strategies successfully reduced the speed standard deviation and average travel time by 28.77% and 7.25% respectively. Moreover, the generalization of the proposed continual learning based VSL approach were evaluated and discussed.

Forgetting to Take Medication, Treatment Adherence and Their Relationship with Viral Load Suppression Among People Living with HIV in the Kilimanjaro Region, Tanzania.

Background: Antiretroviral therapy (ART) adherence is crucial for virological suppression and positive treatment outcomes among people living with HIV (PLHIV), but remains a challenge in ensuring patients achieve and sustain viral load suppression. Despite the recommended use of digital tools medications uptake reminders, the contribution of forgetting to take medication is unknown. This study investigated the contribution of forgetting to take medication on the total missed medication and its effects on detectable viral load (VL). Methods: This mixed-method research was conducted among children, adolescents, pregnant, and breastfeeding women living with HIV on ART in northern Tanzania. Forgetting to take medication constituted reporting to have missed medication due to forgetfulness. A multivariable logistic regression model was used to estimate the adjusted odds ratio (AOR) with a 95% confidence interval (CI) to determine the contribution of forgetting medication intakes on total missed medication and other factors associated with having a detectable VL. Results: Of 427 respondents, 33.3% were children, 33.4% adolescents, and 33.3% pregnant and breastfeeding women, whose median age (interquartile range) was 9 (7-12), 18 (16-18), and 31 (27-36) years, respectively. Ninety-two (22.3%) reported missing medication over the past month, of which 72 (17.9%) was due to forgetting. Forgetting to take medication (AOR: 1.75 95% CI: 1.01-3.06) and being on second-line regimen (AOR: 2.89 95% CI: 1.50-5.55) increased the chances of a detectable VL, while females had lower chances of detectable VL (AOR: 0.62 95% CI: 0.41-0.98). The themes on the reasons for forgetting to take medication from qualitative results included being busy with work and the importance of reminders. Conclusion: Forgetting to take medication is common among PLHIV and an important predictor of a detectable VL. This calls for the use of automated short message services (SMS) reminders or Digital Adherence Tools with reminders to improve and promote good ART adherence among PLHIV.

[Memory impairments in temporal lobe epilepsy].

Temporal lobe epilepsy is known to present with various cognitive impairments, among which memory deficits are frequently reported by patients. Memory deficits can be classified into two types: classical hippocampal amnesia, which is characterized by abnormalities detected in neuropsychological assessments, and atypical memory deficits, such as accelerated long-term amnesia and autobiographical memory impairment, which cannot be identified using standard testing methods. These deficits are believed to arise from a complex interplay among structural brain abnormalities, interictal epileptic discharges, pharmacological factors, and psychological states. While fundamental treatments are limited, there are opportunities for interventions such as environmental adjustments and rehabilitation. This review article aims to provide a comprehensive overview of the types, underlying pathophysiology, and intervention methods for memory disorders observed in patients with temporal lobe epilepsy.

Working memory forgetting: Bridging gaps between human and animal studies.

The causes of forgetting in working memory (WM) remain a source of debate in cognitive psychology, partly because it has always been challenging to probe the complex neural mechanisms that govern rapid cognitive processes in humans. In this review, we argue that neural, and more precisely animal models, provide valuable tools for exploring the precise mechanisms of WM forgetting. First, we discuss theoretical perspectives concerning WM forgetting in humans. Then, we present neuronal correlates of WM in animals, starting from the initial evidence of delay activity observed in the prefrontal cortex to the later synaptic theory of WM. In the third part, specific theories of WM are discussed, including the notion that silent versus non-silent activity is more consistent with the processes of refreshing and decay proposed in human cognitive models. The review concludes with an exploration of the relationship between long-term memory and WM, revealing connections between these two forms of memory through the long-term synaptic hypothesis, which suggests that long-term storage of interference can potentially disrupt WM.

LAM Test: A New Cognitive Marker for Early Detection in Preclinical Alzheimer's Disease.

Background: With the arrival of disease-modifying treatments, it is mandatory to find new cognitive markers that are sensitive to Alzheimer's disease (AD) pathology in preclinical stages. Objective: To determine the utility of a newly developed Learning and Associative Memory face test: LAM test. This study examined the relationship between AD cerebrospinal fluid (CSF) biomarkers and performance on LAM test, and assessed its potential clinical applicability to detect subtle changes in cognitively healthy subjects at risk for AD. Methods: We studied eighty cognitively healthy volunteers from the Valdecilla cohort. 61% were women and the mean age was 67.34 years (±6.416). All participants underwent a lumbar puncture for determination of CSF biomarkers and an extensive neuropsychological assessment, including performance on learning and associative memory indices of the LAM-test after 30 min and after 1 week, and two classic word lists to assess verbal episodic memory: the Rey Auditory Verbal Learning Test (RAVLT) and the Free and Cued Selective Reminding Test (FCSRT). We analyzed cognitive performance according to amyloid status (A+ versus A-) and to ATN model (A-T-N-; A+T-N-; A+T+N-/A+T+N+). Results: Performance on the LAM-test was significantly correlated with CSF Aß ratio. A+ participants performed worse on both learning (mean difference = 2.19, p = 0.002) and memory LAM measures than A- (mean difference = 2.19, p = 0.004). A decline in performance was observed along the Alzheimer's continuum, with significant differences between ATN groups. Conclusions: Our findings suggest that LAM test could be a useful tool for the early detection of subjects within the AD continuum, outperforming classical memory tests.

Task Termination Triggers Spontaneous Removal of Information From Visual Working Memory.

Working memory (WM) is a goal-directed memory system that actively maintains a limited amount of task-relevant information to serve the current goal. By this definition, WM maintenance should be terminated after the goal is accomplished, spontaneously removing no-longer-relevant information from WM. Past studies have failed to provide direct evidence of spontaneous removal of WM content by allowing participants to engage in a strategic reallocation of WM resources to competing information within WM. By contrast, we provide direct neural and behavioral evidence that visual WM content can be largely removed less than 1 s after it becomes obsolete, in the absence of a strategic allocation of resources (total N = 442 adults). These results demonstrate that visual WM is intrinsically a goal-directed system, and spontaneous removal provides a means for capacity-limited WM to keep up with ever-changing demands in a dynamic environment.

Federated learning using model projection for multi-center disease diagnosis with non-IID data.

Multi-center disease diagnosis aims to build a global model for all involved medical centers. Due to privacy concerns, it is infeasible to collect data from multiple centers for training (i.e., centralized learning). Federated Learning (FL) is a decentralized framework that enables multiple clients (e.g., medical centers) to collaboratively train a global model while retaining patient data locally for privacy. However, in practice, the data across medical centers are not independently and identically distributed (Non-IID), causing two challenging issues: (1) catastrophic forgetting at clients, i.e., the local model at clients will forget the knowledge received from the global model after local training, causing reduced performance; and (2) invalid aggregation at the server, i.e., the global model at the server may not be favorable to some clients after model aggregation, resulting in a slow convergence rate. To mitigate these issues, an innovative Federated learning using Model Projection (FedMoP) is proposed, which guarantees: (1) the loss of local model on global data does not increase after local training without accessing the global data so that the performance will not be degenerated; and (2) the loss of global model on local data does not increase after aggregation without accessing local data so that convergence rate can be improved. Extensive experimental results show that our FedMoP outperforms state-of-the-art FL methods in terms of accuracy, convergence rate and communication cost. In particular, our FedMoP also achieves comparable or even higher accuracy than centralized learning. Thus, our FedMoP can ensure privacy protection while outperforming centralized learning in accuracy and communication cost.

Evidence-Based Reconstruction of Memories of a Disdainful Hookup: Identifying Risk Factors and Preventing Further Victimization.

Autobiographical memories of close relationships have been shown to have strong influence in health and life. Yet, there is no research published about longitudinal memory reconstruction of violent sporadic relationships while reading and discussing scientific evidence on gender violence victimization. This article presents a novel case of the reconstruction throughout time of the memory of a disdainful hookup experienced by a young woman. The victim's diary and an interview were the sources of data collection. The analytical categories were developed in dialogue with the participant. The results indicate that, as the subject learned scientific evidence on gender violence in sporadic relationships, she progressively recalled details of the episode that she had self-censored before, became aware of the very violent nature of the hookup, rejected the relationship, and freed her desire for satisfactory romantic relationships.

Memory from nonsense syllables to novels: A survey of retention.

Memory has been the subject of scientific study for nearly 150 years. Because a broad range of studies have been done, we can now assess how effective memory is for a range of materials, from simple nonsense syllables to complex materials such as novels. Moreover, we can assess memory effectiveness for a variety of durations, anywhere from a few seconds up to decades later. Our aim here is to assess a range of factors that contribute to the patterns of retention and forgetting under various circumstances. This was done by taking a meta-analytic approach that assesses performance across a broad assortment of studies. Specifically, we assessed memory across 256 papers, involving 916 data sets (e.g., experiments and conditions). The results revealed that exponential-power, logarithmic, and linear functions best captured the widest range of data compared with power and hyperbolic-power functions. Given previous research on this topic, it was surprising that the power function was not the best-fitting function most often. Contrary to what would be expected, a substantial amount of data also revealed either stable memory over time or improvement. These findings can be used to improve our ability to model and predict the amount of information retained in memory. In addition, this analysis of a large set of memory data provides a foundation for expanding behavioral and neuroimaging research to better target areas of study that can inform the effectiveness of memory.