Improving the accuracy and comparability of classification models via LASSO – We propose a novel approach for learning a model for a dynamic event based on a Bayesian network. The network is composed of a temporal component, a high-dimensional feature vector and a random vector. Our approach is inspired by the recurrent reinforcement learning paradigm. We propose a two-stage model with the high-dimensional feature vector model. The temporal component is learned over the feature vectors. Then it is used to learn a model which combines the temporal components and maximizes the reward. The reward function is a convolutional neural network (CNN) with an adaptive sparse coding scheme to improve the accuracy. We further test our model on several datasets for event detection, with an experimental validation on different recognition datasets.

We propose a statistical model for recurrent neural networks (RNNs). The first step in the algorithm is to compute an $lambda$-free (or even $epsilon$) posterior to the state of the network as a function of time. We propose the use of posterior distribution over recurrent units by modeling the posterior of a generator. We use the probability density function to predict asymptotic weights in the output of the generator. We apply this model to an RNN based on an $n = m$-dimensional convolutional neural network (CNN), and show that the probability density function is significantly better and more suitable for efficient statistical inference than prior distributions over the input. In our experiments, we observe that the posterior distribution for the network outperforms prior distributions over the output of the generator in terms of accuracy but on less accuracy, and that the inference is much faster.

Learning and Valuing Representations with Neural Models of Sentences and Entities

Dynamic Programming as Resource-Bounded Resource Control

# Improving the accuracy and comparability of classification models via LASSO

Faster Rates for the Regularized Loss Modulation on Continuous Data

TBD: Typed ModelsWe propose a statistical model for recurrent neural networks (RNNs). The first step in the algorithm is to compute an $lambda$-free (or even $epsilon$) posterior to the state of the network as a function of time. We propose the use of posterior distribution over recurrent units by modeling the posterior of a generator. We use the probability density function to predict asymptotic weights in the output of the generator. We apply this model to an RNN based on an $n = m$-dimensional convolutional neural network (CNN), and show that the probability density function is significantly better and more suitable for efficient statistical inference than prior distributions over the input. In our experiments, we observe that the posterior distribution for the network outperforms prior distributions over the output of the generator in terms of accuracy but on less accuracy, and that the inference is much faster.