LIBIRWLS - A parallel IRWLS library to solve SVMs and semiparametric SVMs

LIBIRWLS - A parallel IRWLS library to solve SVMs and semiparametric SVMs

A parallel IRWLS library to solve SVMs and semiparametric SVMs

Artificial Intelligence

  • 0 Collaborators

  • 68 Followers

Description

SVMs are a very popular machine learning technique because they can easily create non-linear solutions by transforming the input space onto a high dimensional one where a kernel function can compute the inner product of a pair vectors. Thanks to this ability, they offer a good compromise between complexity and performance in many applications.

SVMs have two main limitations. The first problem is related to their non-parametric nature. The complexity of the classifier is not limited and depends on the number of Support Vectors (SVs) after training. If the number of SVs is very large we may obtain a very slow classifier when processing new samples. The second problem is the run time associated to the training procedure that may be excessive for large datasets.

To face these problems, we can make use of parallel computing, thus reducing the run time of the training procedure or we can use semi-parametric approximations than can limit the complexity of the model in advance, which directly implies a faster classifier.

The above situation motivated us to develop "LIBIRWLS", an integrated library based on a parallel implementation of the IRWLS procedure to solve non-linear SVMs and semi-parametric SVMs. This library is implemented in C, supports a wide range of platforms and also provides detailed information about its programming interface and dependencies.

It implements the functions to run two different algorithms:

Parallel Iterative Re-Weighted Least Squares: A Parallel SVM solver based on the IRWLS algorithm.

Parallel Semi-parametric Iterative Re-Weighted Least Squares: A Parallel Semiparametric SVMs solver based on the IRWLS algorithm.

Links

Web Page

Github

Medium pp

Avideep M. updated status

Medium pp

Avideep Mukherjee

I am currently working on kNN classifiers. I want to know a simple case where weighted kNN classifier outperforms kNN classifier. Any help would be appreciated.

Default user avatar

AMIRSINA T. created project Lip Reading - Cross Audio-Visual Recognition using 3D Convolutional Neural Networks

Medium 1f56e0df 5afe 42cb 8546 bd16cc283223

Lip Reading - Cross Audio-Visual Recognition using 3D Convolutional Neural Networks

Audio-visual recognition (AVR) has been considered as a solution for speech recognition tasks when the audio is corrupted, as well as a visual recognition method used for speaker verification in multispeaker scenarios. The approach of AVR systems is to leverage the extracted information from one modality to improve the recognition ability of the other modality by complementing the missing information. The essential problem is to find the correspondence between the audio and visual streams, which is the goal of this paper. We propose the use of a coupled 3D convolutional neural network (3D CNN) architecture that can map both modalities into a representation space to evaluate the correspondence of audio-visual streams using the learned multimodal features. The proposed architecture will incorporate both spatial and temporal information jointly to effectively find the correlation between temporal information for different modalities. By using a relatively small network architecture and much smaller dataset for training, our proposed method surpasses the performance of the existing similar methods for audio-visual matching, which use 3D CNNs for feature representation. We also demonstrate that an effective pair selection method can significantly increase the performance. The proposed method achieves relative improvements over 20% on the equal error rate and over 7% on the average precision in comparison to the state-of-the-art method.

Medium yoshikawa.chinatsu.full.620099

Yuanzheng C. created project ColorEvangelist

Medium e232d9bd aae4 49c3 a274 a7c0f6a1d912

ColorEvangelist

Sketch/Line Art colorization is a time consuming process, The automation of the process requires not just simple boundary detection but also semantical feature identification, user interaction and shading(which is not a problem for grey image colorization). We propose an deep end-to-end trainable colorization model that meanwhile small in size and has a almost-real-time performance. It identifies semantic features from the sketch and inpaint it with/without user interaction with realistic shading.

Medium 17884422 394019487650747 6366979724171094013 n

Alexander L. created project Analyzing Radio Telescope Array Big Data Using Intel DevCloud TensorFlow

Medium 6a30d196 dbf0 4974 890a fe91f432c489

Analyzing Radio Telescope Array Big Data Using Intel DevCloud TensorFlow

For decades, institutional data on radio telescope arrays have accumulated in American astronomical research institutions.

Radio array telescopes like SETI at Home and the NRAO Very Large Array have over time, amassed big, not just large, big data sets of astronomical data concerning the cosmos. Data points such as polarization, radio wave frequency, magnitude, longitude/latitude, and date/time help astronomers pinpoint and understand stars.

We seek to use TensorFlow integrated into Intel DevCloud to build a TensorFlow Machine Learning model which better interprets and analyzes radio telescope array big data and also to utilize machine vision to further enhance radio images created by astronomical big data.

Mentored by Professor Micheal Strong at Cosumnes River College, Sacramento CA. 2018

See More

No users to show at the moment.

Bigger mifoto
  • Projects 1
  • Followers 5

Roberto Diaz Morales

I am currently the leader of the Data Science research line at Treelogic where I take part in many European H2020 research projects. My research interest include large scale Machine Learning, Parallelization and Deep Learning techniques. I obtained my PhD and Master Degree from the University Carlos III in Madrid where I worked under the supervision of Ángel Navia Vázquez. My Ph.D. research consisted in the parallelization of kernel methods, a family of of Machine Learning algorithms like Support Vector Machines (SVMs) or Gaussian Processes (GPs). My hobbies include taking part in Machine Learning competitions, such as those organized by Kaggle where I have received prizes in many competitions and I have reached the top 100 of the world ranking of data scientist.

See More