Operator-valued Kernels for Learning from Functional Response Data
Authors: Hachem Kadri, Emmanuel Duflos, Philippe Preux, Stéphane Canu, Alain Rakotomamonjy, Julien Audiffren
JMLR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In Section 7, we illustrate the performance of our approach through speech and audio processing experiments. ... In this experimental section, we essentially aim at illustrating the potential of adopting a functional data analysis perspective for learning multi-output functions when the data are curves. ... Table 3 reports average RSSE results obtained using the hierarchical ε-SVR algorithm (Mitra et al., 2009), the multi-task kernel method (Kadri et al., 2011a) after smoothing the estimated VT trajectories using a Kalman filter as described in Mitra et al. (2009), and the functional operator-valued kernel based approach. |
| Researcher Affiliation | Academia | Hachem Kadri EMAIL Aix-Marseille Universit e, LIF (UMR CNRS 7279) F-13288 Marseille Cedex 9, France ... Emmanuel Duflos EMAIL Ecole Centrale de Lille, CRISt AL (UMR CNRS 9189) 59650 Villeneuve d Ascq, France ... Philippe Preux EMAIL Universit e de Lille, CRISt AL (UMR CNRS 9189) 59650 Villeneuve d Ascq, France ... St ephane Canu EMAIL INSA de Rouen, LITIS (EA 4108) 76801, St Etienne du Rouvray, France ... Alain Rakotomamonjy EMAIL Universit e de Rouen, LITIS (EA 4108) 76801, St Etienne du Rouvray, France ... Julien Audiffren EMAIL ENS Cachan, CMLA (UMR CNRS 8536) 94235 Cachan Cedex, France |
| Pseudocode | Yes | Algorithm 1 L2-Regularized Function-valued Function Learning Algorithm |
| Open Source Code | No | The paper does not provide explicit statements about the release of open-source code or links to a code repository for the described methodology. |
| Open Datasets | Yes | Most current research on acoustic-to-articulatory inversion focuses on learning Electromagnetic Articulography (EMA) trajectories from acoustic parameters and frequently uses the MOCHA fsew0 data set as training and test data (Richmond, 2002). ... acoustic-articulatory database is generated by the TADA model (Nam et al., 2004) which is a computational implementation of articulatory phonology. The generated data set consists of acoustic signals for 416 words chosen from the Wisconsin X-ray microbeam data (Westbury et al., 1994) and corresponding Vocal Tract (VT) trajectories sampled at 5 ms. ... The major part of the sound samples used in the recognition experiments is taken from two sound libraries (Leonardo Software; Real World Computing Paternship, 2000). URL http://tosa.mri.co.jp/sounddb/indexe.htm. Leonardo Software. http://www.leonardosoft.com. |
| Dataset Splits | Yes | Table 4: Classes of sounds and number of samples in the database used for performance evaluation. ... Classes Number Train Test Total. ... Human screams C1 40 25 65. Gunshots C2 36 19 55. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, or cloud instance types) used for running its experiments. |
| Software Dependencies | No | The paper mentions methods like 'hierarchical ε-SVR' and 'Kalman filter' but does not provide specific version numbers for any software libraries or tools used in their implementation. |
| Experiment Setup | No | The paper states, 'In order to choose the regularization parameter λ and the number of eigenfunctions κ that guarantee optimal solutions, one may use the cross-validation score based on the one-curve-leave-out prediction error (Rice and Silverman, 1991),' but it does not provide the specific values of these parameters or other training settings used in the experiments. |