Research Statement Samory Kpotufe
نویسنده
چکیده
Given the range and variety of modern applications of pattern-recognition, it is becoming increasingly important to design procedures which can self-tune to the statistical difficulty of the learning problem at hand, within the computational constraints of modern data sizes. These so-called adaptive procedures must automatically adjust their behavior to achieve performance on par with the best alternatives for the given problem, and should do so in an efficient way.
منابع مشابه
Research Statement Samory Kpotufe General Research Motivation
I work in Machine Learning, a field at the intersection of Statistics and Computer Science. The field is concerned with the design of learning algorithms which automatically adapt to new scenarios by learning from past observations. Application domains abound in engineering and scientific fields; some examples are speech recognition, computer vision, genomics, medical diagnosis, web mining, and...
متن کاملOptimal rates for k-NN density and mode estimation
We present two related contributions of independent interest: (1) high-probability finite sample rates for k-NN density estimation, and (2) practical mode estimators – based on k-NN – which attain minimax-optimal rates under surprisingly general distributional conditions.
متن کاملEscaping the Curse of Dimensionality with a Tree-based Regressor
We present the first tree-based regressor whose convergence rate depends only on the intrinsic dimension of the data, namely its Assouad dimension. The regressor uses the RPtree partitioning procedure, a simple randomized variant of k-d trees.
متن کاملConsistency of Causal Inference under the Additive Noise Model
We analyze a family of methods for statistical causal inference from sample under the so-called Additive Noise Model. While most work on the subject has concentrated on establishing the soundness of the Additive Noise Model, the statistical consistency of the resulting inference methods has received little attention. We derive general conditions under which the given family of inference methods...
متن کاملGradient Weights help Nonparametric Regressors
where An,i(X) ≡ we are confident in both estimates fn,h(X ± tei). •Fast preprocessing, and online: just 2 estimates of fn,h at X . Metric learning optimizes over a space of possible metrics. •Only 2× 2× 1 param. search grid to adapt to d dimensions. KR with bandwidths {hi}d1 needs d× d param. search grid. •General: preprocessing for any distance-based regressor. Other methods apply to particula...
متن کامل