Statistical transfer learning: A review and some extensions to statistical process control

Tsung, Fugee; Zhang, Ke; Cheng, Longwei; Song, Zhenli

doi:10.1080/08982112.2017.1373810

Cited by 55 publications

(21 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When the training set and the testing set comprise images of similar context, it is possible to take the features that are closer to the output layer. However, when the sets are from completely different domains, it is recommended to a take the features from pooling layers residing closer to the middle of the network, which represent basic shapes such as lines, circles, and trends [ 37 ]. We preferred using a single-neuron model to more complicated multi-layered architectures due to the size of the training set (for detailed results, see Appendix E ).…”

Section: Resultsmentioning

confidence: 99%

Level-K Classification from EEG Signals Using Transfer Learning

Mizrahi

Zuckerman

Laufer

2021

Sensors

View full text Add to dashboard Cite

Tacit coordination games are games in which communication between the players is not allowed or not possible. In these games, the more salient solutions, that are often perceived as more prominent, are referred to as focal points. The level-k model states that players’ decisions in tacit coordination games are a consequence of applying different decision rules at different depths of reasoning (level-k). A player at Lk=0 will randomly pick a solution, whereas a Lk≥1 player will apply their strategy based on their beliefs regarding the actions of the other players. The goal of this study was to examine, for the first time, the neural correlates of different reasoning levels in tacit coordination games. To that end, we have designed a combined behavioral-electrophysiological study with 3 different conditions, each resembling a different depth reasoning state: (1) resting state, (2) picking, and (3) coordination. By utilizing transfer learning and deep learning, we were able to achieve a precision of almost 100% (99.49%) for the resting-state condition, while for the picking and coordination conditions, the precision was 69.53% and 72.44%, respectively. The application of these findings and related future research options are discussed.

show abstract

Section: Resultsmentioning

confidence: 99%

Level-K Classification from EEG Signals Using Transfer Learning

Mizrahi

Zuckerman

Laufer

2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Besides solid color shirts, the aesthetic perception for shirts of other patterns also needs to be investigated using the transfer learning method [44], based on the developed model, including two-color, multi-color, and one-color gradient shirts, as shown in Figure 1.…”

Section: Research Limitationsmentioning

confidence: 99%

Perception and Preference Analysis of Fashion Colors: Solid Color Shirts

2019

View full text Add to dashboard Cite

When it comes to pollution, we do not usually think about the clothes we wear, but the clothing industry is really endangering our planet. The market economy has transferred the decision-making power of the garment industry from enterprises to consumers. To make the fashion industry sustainable, in addition to technological innovation, it is also necessary to conduct research on the service objects of the industry. Consumer clothing preference research is an important part of the sustainable development of the clothing industry, and it will also have an impact on environmental and design sustainability. Hence, a psychophysical experiment based on solid color shirts is carried out to analyze people’s perceptions and preferences concerning fashion colors, including the aesthetic differences and similarities between males and females, and establish a hierarchical feed-forward model of color preferences relating to solid color shirts. Firstly, 480 colors of solid shirts from different clothing brands were collected, and the mean shift clustering algorithm was used to classify them into 19 clusters in the CIELAB color space. Secondly, another 22 solid colors, combined with the 19 colors of the cluster centers, formed a solid color scheme. Thirdly, 41 solid male and female shirts and fabrics were simulated as stimuli in three dimensions, and they were presented on a calibrated computer display. The simulations were assessed by 34 observers (consisting of 17 males and 17 females) in terms of 11 semantic scales, including cold/warm, heavy/light, passive/active, dirty/clean, tense/relaxed, plain/gaudy, traditional/modern, masculine/feminine, slim-look/fat-look, hard-to-match/easy-to-match, and dislike/tike. The experimental results demonstrated that the hard-to-match/easy-to-match response was found to be highly correlated with dislike/like. Furthermore, the response of the females concerning hard-to-match/easy-to-match had a strong correlation with two adjective pairs (plain/gaudy and slim-look/fat-look), while that of the males also had a strong correlation with two adjective pairs (plain/ gaudy and masculine/feminine). Finally, a hierarchical feed-forward model of aesthetic perception for solid color shirts was established to predict the shirt preference degree. These findings could be used to develop a more robust and comprehensive theory of fashion color preferences and provide a reference for the design of solid color shirts. A more comprehensive color preference theory is not only an effective tool to solve the problem of pollution in the clothing industry, but also an important theoretical basis for the “sustainable design” of clothing, which is of great significance to the sustainable development of the clothing industry.

show abstract

“…For example, in [72] the knowledge for solving a simple version of a problem is transferred to a more complex one -transfer learning from 2D to 3D mountain car problem; transfer learning from a Mario game without enemies to a Mario game with enemies. More in-depth reviews of transfer learning techniques are provided in [71,73,74].…”

Section: Transfer Learningmentioning

confidence: 99%

Machine Autonomy: Definition, Approaches, Challenges and Research Gaps

Ezenkwu

Starkey

2019

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

The processes that constitute the designs and implementations of AI systems such as self-driving cars, factory robots and so on have been mostly hand-engineered in the sense that the designers aim at giving the robots adequate knowledge of its world. This approach is not always efficient especially when the agent's environment is unknown or too complex to be represented algorithmically. A truly autonomous agent can develop skills to enable it to succeed in such environments without giving it the ontological knowledge of the environment a priori. This paper seeks to review different notions of machine autonomy and presents a definition of autonomy and its attributes. The attributes of autonomy as presented in this paper are categorised into low-level and high-level attributes. The low-level attributes are the basic attributes that serve as the separating line between autonomous and other automated systems while the high-level attributes can serve as a taxonomic framework for ranking the degrees of autonomy of any system that has passed the low-level autonomy. The paper reviews some AI techniques as well as popular AI projects that focus on autonomous agent designs in order to identify the challenges of achieving a true autonomous system and suggest possible research directions.

show abstract

Statistical transfer learning: A review and some extensions to statistical process control

Cited by 55 publications

References 35 publications

Level-K Classification from EEG Signals Using Transfer Learning

Level-K Classification from EEG Signals Using Transfer Learning

Perception and Preference Analysis of Fashion Colors: Solid Color Shirts

Machine Autonomy: Definition, Approaches, Challenges and Research Gaps

Contact Info

Product

Resources

About