Probabilities for SV Machines

Smola, Alexander J.; Bartlett, Peter; Schölkopf, Bernhard; Schuurmans, Dale

doi:10.7551/mitpress/1113.003.0008

Cited by 750 publications

(85 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We first selected five baseline machine learning methods to predict microbial population adaptation, including RF with the squared error criterion, extreme gradient boosting (XGBoost) with the squared error loss function, gradient boosting regression tree (GBRT) with the squared error loss function, support vector machine (SVM) with a radial basis function kernel and multilayer perceptron (MLP). The predictive performances were compared using a tenfold cross-validation test (Chen & Guestrin, 2016;Friedman, 2002;He et al, 2015;Platt, 2000). For these classifiers, we utilized the grid search method to select the optimal parameters.…”

Section: Ensemble Machine Learning Model Developmentmentioning

confidence: 99%

Ensemble learning model identifies adaptation classification and turning points of river microbial communities in response to heatwaves

Qu,

Xu,

Kang

et al. 2023

Global Change Biology

View full text Add to dashboard Cite

Heatwaves are a global issue that threaten microbial populations and deteriorate ecosystems. However, how river microbial communities respond to heatwaves and whether and how high temperatures exceed microbial adaptation remain unclear. In this study, we proposed four types of pulse temperature‐induced microbial responses and predicted the possibility of microbial adaptation to high temperature in global rivers using ensemble machine learning models. Our findings suggest that microbial communities in parts of South American (e.g., Brazil and Chile) and Southeast Asian (e.g., Vietnam) countries are likely to change due to heatwave disturbance from 25 to 37°C for consecutive days. Furthermore, the microbial communities in approximately 48.4% of the global river gauge stations are prone to fast stress inadaptation, with approximately 76.9% of these stations expected to exceed microbial adaptation after heatwave disturbances. If emissions of particulate matter with sizes not more than 2.5 μm (PM2.5, an indicator of human activities) increase by twofold, the number of global rivers associated with the fast stress adaptation type will decrease by ~13.7% after heatwave disturbances. Understanding microbial responses is crucially important for effective ecosystem management, especially for fragile and sensitive rivers facing heatwave events.

show abstract

Section: Ensemble Machine Learning Model Developmentmentioning

confidence: 99%

Ensemble learning model identifies adaptation classification and turning points of river microbial communities in response to heatwaves

Qu,

Xu,

Kang

et al. 2023

Global Change Biology

View full text Add to dashboard Cite

show abstract

“…For predictions based on SVM models, instead of using a boolean output value indicating 1D or non-1D material based on the location of a feature vector relative to the separating hyperplane, we generate a continuous classification between 0 and 1 using Platt scaling. 27 For random forest models, a probability can likewise be determined by the percentage of individual decision trees (100 used) where the individual tree predicts a composition to be 1D. The relevant threshold of the probability prediction can be chosen not only based on the precision−recall curve for the test set but also on the number of predicted positive 1D materials when the model is used for screening.…”

Section: The Journal Of Physical Chemistrymentioning

confidence: 99%

Machine-Learning-Driven Expansion of the 1D van der Waals Materials Space

Zhu,

Antoniuk,

Wright

et al. 2023

J. Phys. Chem. C

View full text Add to dashboard Cite

One-dimensional (1D) van der Waals (vdW) materials display electronic and magnetic transport properties that make them uniquely suited as interconnect materials and for low-dimensional optoelectronic applications. However, there are only around 700 1D vdW structures in general materials databases, making database curation approaches ineffective for 1D discovery. Here, we utilize machine-learning techniques to discover 1D vdW compositions that have not yet been synthesized. Our techniques go beyond discovery efforts involving elemental substitutions and instead start with a composition space of 4741 binary and 392,342 ternary formulas. We predict up to 3000 binary and 10,000 ternary 1D compounds and further classify them by expected magnetic and electronic properties. Our model identifies MoI3, a material we experimentally confirm to exist with wire-like subcomponents and exotic magnetic properties. More broadly, we find several chalcogen-, halogen-, and pnictogen-containing compounds expected to be synthesizable using chemical vapor deposition and chemical vapor transport.

show abstract

“…{g} vs {m, s}, {m} vs {s, g}, {s} vs {m, g}) and three binary SVM classifiers for the vowels ({a} vs {i, u}, {i} vs {a, u}, {u} vs {a, i}), aggregating the results 39,40 . The unnormalised scores that this produces can be mapped to interpretable posterior probabilities for consonants and for vowels 41,42 . The input features to the SVMs were the same as for the DNNs (i.e.…”

Section: Support Vector Machinesmentioning

confidence: 99%

A note on decoding elicited and self-generated inner speech

Jones

Voets

2021

Preprint

View full text Add to dashboard Cite

A recent result shows that inner speech can, with proper care, be decoded to the same high-level of accuracy as articulated speech. This relies, however, on neural data obtained while subjects perform elicited tasks, such as covert reading and repeating, whereas a neural speech prosthetic will require the decoding of inner speech that is self-generated. Prior work has, moreover, emphasised differences between these two kinds of inner speech, raising the question of how well a decoder optimised for one will generalise to the other. In this study, we trained phoneme-level decoders on an atypically large, elicited inner speech dataset, previously acquired using 7T fMRI in a single subject. We then acquired a second self-generated inner speech dataset in the same subject. Although the decoders were trained exclusively on neural recordings obtained during elicited inner speech, they predicted unseen phonemes accurately in both elicited and self-generated test conditions, illustrating the viability of zero-shot task transfer. This has significant practical importance for the development of a neural speech prosthetic, as labelled data is far easier to acquire at scale for elicited than for self-generated inner speech. Indeed, elicited tasks may be the only option for acquiring labelled data in critical patient populations who cannot control their vocal articulators.

show abstract

Probabilities for SV Machines

Cited by 750 publications

References 0 publications

Ensemble learning model identifies adaptation classification and turning points of river microbial communities in response to heatwaves

Ensemble learning model identifies adaptation classification and turning points of river microbial communities in response to heatwaves

Machine-Learning-Driven Expansion of the 1D van der Waals Materials Space

A note on decoding elicited and self-generated inner speech

Contact Info

Product

Resources

About