You will design and implement state-of-the-art methods for both supervised and unsupervised learning, with a focus on sensor data such as gyroscope and accelerometer streams. Your knowledge of signal processing allows you to apply the necessary pre-processing steps such as band pass filtering, down sampling while anti-aliasing, Fourier or Cepstrum coefficient extraction and spectrogram modeling. You have experience with temporal modeling techniques, both for discrete state spaces (e.g. You have provable experience with generative (e.g. You have a strong theoretical and mathematical background and are able to reason about machine learning peculiarities in order to answer questions such as: Is a Bayesian classifier with Gaussian likelihoods and priors the same as a Euclidean distance classifier if equal and diagonal covariance matrices are used?

Support Vector Machine (SVM) is powerful classification technique based on the idea of structural risk minimization. Use of kernel function enables curse of dimensionality to be addressed. However, proper kernel function for certain problem is dependent on specific dataset and as such there is no good method on choice of kernel function. In this paper, SVM is used to build empirical models of currency crisis in Argentina. An estimation technique is developed by training model on real life data set which provides reasonably accurate model outputs and helps policy makers to identify situations in which currency crisis may happen. The third and fourth order polynomial kernel is generally best choice to achieve high generalization of classifier performance. SVM has high level of maturity with algorithms that are simple, easy to implement, tolerates curse of dimensionality and good empirical performance. The satisfactory results show that currency crisis situation is properly emulated using only small fraction of database and could be used as an evaluation tool as well as an early warning system. To the best of knowledge this is the first work on SVM approach for currency crisis evaluation of Argentina.

Digital media tend to combine text and images to express richer information, especially on image hosting and online shopping websites. This trend presents a challenge in understanding the contents from different forms of information. Features representing visual information are usually sparse in high dimensional space, which makes the learning process intractable. In order to understand text and its related visual information, we present a new graphical model-based approach to discover more meaningful information in rich media. We extend the standard Latent Dirichlet Allocation (LDA) framework to learn in high dimensional feature spaces.