Online-adaptive classification and regression network with sample-efficient meta learning for long-term tracking

2021 ◽  
pp. 104181
Author(s):  
Lang Yu ◽  
Huanlong Zhang ◽  
Junyang Yu ◽  
Baojun Qiao
Keyword(s):  
Adaptive Classification ◽  
Meta Learning ◽  
Classification And Regression

scholarly journals Towards Fluid Machine Intelligence: Can We Make a Gifted AI?

2019 ◽  
Vol 33 ◽  
pp. 9760-9764
Author(s):  
Ian Davidson ◽  
Peter B. Walker
Keyword(s):  
School Districts ◽  
Fluid Intelligence ◽  
Machine Intelligence ◽  
Long Term Memory ◽  
Specific Knowledge ◽  
Current Form ◽  
Crystallized Intelligence ◽  
Meta Learning ◽  
Gifted Education Programs

Most applications of machine intelligence have focused on demonstrating crystallized intelligence. Crystallized intelligence relies on accessing problem-specific knowledge, skills and experience stored in long term memory. In this paper, we challenge the AI community to design AIs to completely take tests of fluid intelligence which assess the ability to solve novel problems using problem-independent solving skills. Tests of fluid intelligence such as the NNAT are used extensively by schools to determine entry into gifted education programs. We explain the differences between crystallized and fluid intelligence, the importance and capabilities of machines demonstrating fluid intelligence and pose several challenges to the AI community, including that a machine taking such a test would be considered gifted by school districts in the state of California. Importantly, we show existing work on seemingly related fields such as transfer, zero-shot, life-long and meta learning (in their current form) are not directly capable of demonstrating fluid intelligence but instead are task-transductive mechanisms.

Analytical procedures for evaluating factors that affect joint faulting for jointed plain concrete pavements using the Long-Term Pavement Performance database

2006 ◽  
Vol 33 (10) ◽  
pp. 1279-1286
Author(s):  
Jong-Suk Jung ◽  
Emmanuel B Owusu-Antwi ◽  
Ji-Hwan An
Keyword(s):  
Cluster Analysis ◽  
Regression Tree ◽  
Plain Concrete ◽  
Classification And Regression Tree ◽  
Pavement Performance ◽  
Concrete Pavements ◽  
Analysis Techniques ◽  
Classification And Regression ◽  
Design And Construction

The objective of this study was to identify and quantify design and construction features most important to joint faulting of joint plain concrete pavements. With data obtained from the Long-Term Pavement Performance (LTPP) database, an analysis approach that combined pavement engineering expertise and modern data analysis techniques was to develop guidelines for improved design and construction of Portland cement concrete (PCC) pavement. The approach included typical preliminary analyses, but emphasis was placed on using a series of multivariate data analysis techniques. Discriminant analysis was used to develop models that classify individual pavement into performance groups developed by cluster analysis, which was used to partition the pavements into three distinct groups representing good, normal, and poor performance. These models can be used to classify and evaluate additional or new pavements performance throughout the pavement's design life. To quantify the levels of the key design and construction features that contribute to performance, the classification and regression tree procedure was used to develop tree-based models for performance measure. The analysis approach described was used to develop the guideline on the key design and construction features that can be used by designers to decrease joint faulting of jointed plain concrete pavements (JPCPs).Key words: faulting, Long-Term Pavement Performance (LTPP), jointed plain concrete pavement (JPCP), cluster analysis, discriminant analysis, classification and regression tree (CART) analysis.

Neural networks for long-term earthquake prediction using modified meta-learning

2021 ◽  
pp. 1-14
Author(s):  
Mario Maya ◽  
Wen Yu ◽  
Luciano Telesca
Keyword(s):  
Neural Networks ◽  
Neural Model ◽  
Learning Rates ◽  
Learning Technique ◽  
Meta Learning ◽  
The Mean ◽  
The Stability ◽  
Error Index ◽  
Long Term Prediction

Neural networks have been successfully applied for modeling time series. However, the results of long-term prediction are not satisfied. In this paper, the modified Meta-Learning is applied to the neural model. The normal Meta-Learning is modified by time-varying learning rates and adding a momentum term to improve convergence speed and robustness property. The stability of the learning process is proven. Finally, two experiments are presented to evaluate the proposed method. The first one shows an improvement in earthquakes prediction in the long-term, and the second one is a classical Benchmark problem. In both experiments, the modified Meta-Learning technique minimizes remarkably the mean square error index.

scholarly journals Not so primitive: context-sensitive meta-learning about unattended sound sequences

2013 ◽  
Vol 109 (1) ◽  
pp. 99-105 ◽  
Author(s):  
Juanita Todd ◽  
Alexander Provost ◽  
Lisa R. Whitson ◽  
Gavin Cooper ◽  
Andrew Heathcote
Keyword(s):  
Short Term Memory ◽  
Long Term Memory ◽  
Term Memory ◽  
Context Sensitive ◽  
Multiple Context ◽  
Auditory Scene ◽  
Meta Learning ◽  
Memory Representations ◽  
Evoked Response Potential

Mismatch negativity (MMN), an evoked response potential elicited when a “deviant” sound violates a regularity in the auditory environment, is integral to auditory scene processing and has been used to demonstrate “primitive intelligence” in auditory short-term memory. Using a new multiple-context and -timescale protocol we show that MMN magnitude displays a context-sensitive modulation depending on changes in the probability of a deviant at multiple temporal scales. We demonstrate a primacy bias causing asymmetric evidence-based modulation of predictions about the environment, and we demonstrate that learning how to learn about deviant probability (meta-learning) induces context-sensitive variation in the accessibility of predictive long-term memory representations that underpin the MMN. The existence of the bias and meta-learning are consistent with automatic attributions of behavioral salience governing relevance-filtering processes operating outside of awareness.

scholarly journals Learning from the Past: Continual Meta-Learning with Bayesian Graph Neural Networks

2020 ◽  
Vol 34 (04) ◽  
pp. 5021-5028 ◽  
Author(s):  
Yadan Luo ◽  
Zi Huang ◽  
Zheng Zhang ◽  
Ziwei Wang ◽  
Mahsa Baktashmotlagh ◽  
...  
Keyword(s):  
Neural Networks ◽  
Message Passing ◽  
Shot Classification ◽  
The Past ◽  
Meta Learning ◽  
Error Accumulation ◽  
Effectiveness And Efficiency ◽  
Graph Neural Networks ◽  
Continual Learning

Meta-learning for few-shot learning allows a machine to leverage previously acquired knowledge as a prior, thus improving the performance on novel tasks with only small amounts of data. However, most mainstream models suffer from catastrophic forgetting and insufficient robustness issues, thereby failing to fully retain or exploit long-term knowledge while being prone to cause severe error accumulation. In this paper, we propose a novel Continual Meta-Learning approach with Bayesian Graph Neural Networks (CML-BGNN) that mathematically formulates meta-learning as continual learning of a sequence of tasks. With each task forming as a graph, the intra- and inter-task correlations can be well preserved via message-passing and history transition. To remedy topological uncertainty from graph initialization, we utilize Bayes by Backprop strategy that approximates the posterior distribution of task-specific parameters with amortized inference networks, which are seamlessly integrated into the end-to-end edge learning. Extensive experiments conducted on the miniImageNet and tieredImageNet datasets demonstrate the effectiveness and efficiency of the proposed method, improving the performance by 42.8% compared with state-of-the-art on the miniImageNet 5-way 1-shot classification task.

Sign in / Sign up

Export Citation Format

Share Document