The Expectation--Maximization (EM) algorithm is a simple meta-algorithm that has been used for many years as a methodology for statistical inference when there are missing measurements in the observed data or when the data is composed of observables and unobservables. Its general properties are well studied, and also, there are countless ways to apply it to individual problems. In this paper, we introduce the $em$ algorithm, an information geometric formulation of the EM algorithm, and its extensions and applications to various problems. Specifically, we will see that it is possible to formulate an outlier-robust inference algorithm, an algorithm for calculating channel capacity, parameter estimation methods on probability simplex, particular multivariate analysis methods such as principal component analysis in a space of probability models and modal regression, matrix factorization, and learning generative models, which have recently attracted attention in deep learning, from the geometric perspective.

Information geometry is a framework to analyze statistical inference and machine learning[2]. Geometrically, statistical inference and many machine learning algorithms can be regarded as procedures to find a projection to a model subspace from a given data point. In this paper, we focus on an algorithm to find the projection. Since the projection is given by minimizing a divergence, a common approach to finding the projection is a gradient-based method[6]. However, such an approach is not applicable in some cases. For instance, several attempts to extend the information geometrical framework to nonparametric cases[3, 9, 13, 15], where we need to consider a function space or each data is represented as a point process. In such a case, it is difficult to compute the derivative of divergence that is necessary for gradient-based methods, and in some cases, it is difficult to deal with the coordinate explicitly. Takano et al.[15] proposed a geometrical algorithm to find the projection for nonparametric e-mixture distribution, where the model subspace is spanned by several empirical distributions. The algorithm that is derived based on the generalized Pythagorean theorem only depends on the values of divergences.

We have obtained an integral representation of the shallow neural network that attains the global minimum of its backpropagation (BP) training problem. According to our unpublished numerical simulations conducted several years prior to this study, we had noticed that such an integral representation may exist, but it was not proven until today. First, we introduced a Hilbert space of coefficient functions, and a reproducing kernel Hilbert space (RKHS) of hypotheses, associated with the integral representation. The RKHS reflects the approximation ability of neural networks. Second, we established the ridgelet analysis on RKHS. The analytic property of the integral representation is remarkably clear. Third, we reformulated the BP training as the optimization problem in the space of coefficient functions, and obtained a formal expression of the unique global minimizer, according to the Tikhonov regularization theory. Finally, we demonstrated that the global minimizer is the shrink ridgelet transform. Since the relation between an integral representation and an ordinary finite network is not clear, and BP is convex in the integral representation, we cannot immediately answer the question such as "Is a local minimum a global minimum?" However, the obtained integral representation provides an explicit expression of the global minimizer, without linearity-like assumptions, such as partial linearity and monotonicity. Furthermore, it indicates that the ordinary ridgelet transform provides the minimum norm solution to the original training equation.

The feature map obtained from the denoising autoencoder (DAE) is investigated by determining transportation dynamics of the DAE, which is a cornerstone for deep learning. Despite the rapid development in its application, deep neural networks remain analytically unexplained, because the feature maps are nested and parameters are not faithful. In this paper, we address the problem of the formulation of nested complex of parameters by regarding the feature map as a transport map. Even when a feature map has different dimensions between input and output, we can regard it as a transportation map by considering that both the input and output spaces are embedded in a common high-dimensional space. In addition, the trajectory is a geometric object and thus, is independent of parameterization. In this manner, transportation can be regarded as a universal character of deep neural networks. By determining and analyzing the transportation dynamics, we can understand the behavior of a deep neural network. In this paper, we investigate a fundamental case of deep neural networks: the DAE. We derive the transport map of the DAE, and reveal that the infinitely deep DAE transports mass to decrease a certain quantity, such as entropy, of the data distribution. These results though analytically simple, shed light on the correspondence between deep neural networks and the Wasserstein gradient flows.

Data representation in a stacked denoising autoencoder is investigated. Decoding is a simple technique for translating a stacked denoising autoencoder into a composition of denoising autoencoders in the ground space. In the infinitesimal limit, a composition of denoising autoencoders is reduced to a continuous denoising autoencoder, which is rich in analytic properties and geometric interpretation. For example, the continuous denoising autoencoder solves the backward heat equation and transports each data point so as to decrease entropy of the data distribution. Together with ridgelet analysis, an integral representation of a stacked denoising autoencoder is derived.

We study a population decoding paradigm in which the maximum likelihood inferenceis based on an unfaithful decoding model (UMLI). This is usually the case for neural population decoding because the encoding process of the brain is not exactly known, or because a simplified decoding modelis preferred for saving computational cost. We consider an unfaithful decoding model which neglects the pairwise correlation between neuronal activities, and prove that UMLI is asymptotically efficient whenthe neuronal correlation is uniform or of limited-range. The performance of UMLI is compared with that of the maximum likelihood inference based on a faithful model and that of the center of mass decoding method.It turns out that UMLI has advantages of decreasing the computational complexity remarkablely and maintaining a high-level decoding accuracy at the same time. The effect of correlation on the decoding accuracy is also discussed.

An adaptive online algorithm extending the learning of learning idea is proposed and theoretically motivated. Relying only on gradient flowinformation it can be applied to learning continuous functions or distributions, even when no explicit loss function is given andthe Hessian is not available. Its efficiency is demonstrated for a non-stationary blind separation task of acoustic signals. 1 Introduction Neural networks provide powerful tools to capture the structure in data by learning. Often the batch learning paradigm is assumed, where the learner is given all training examplessimultaneously and allowed to use them as often as desired. In large practical applications batch learning is often experienced to be rather infeasible and instead online learning is employed.

An adaptive online algorithm extending the learning of learning idea is proposed and theoretically motivated. Relying only on gradient flow information it can be applied to learning continuous functions or distributions, even when no explicit loss function is given and the Hessian is not available. Its efficiency is demonstrated for a non-stationary blind separation task of acoustic signals.

A statistical theory for overtraining is proposed. The analysis treats realizable stochastic neural networks, trained with Kullback Leibler loss in the asymptotic case. It is shown that the asymptotic gain in the generalization error is small if we perform early stopping, even if we have access to the optimal stopping time. Considering cross-validation stopping we answer the question: In what ratio the examples should be divided into training and testing sets in order to obtain the optimum performance. In the non-asymptotic region cross-validated early stopping always decreases the generalization error. Our large scale simulations done on a CM5 are in nice agreement with our analytical findings.

A statistical theory for overtraining is proposed. The analysis treats realizable stochastic neural networks, trained with Kullback Leibler loss in the asymptotic case. It is shown that the asymptotic gain in the generalization error is small if we perform early stopping, evenif we have access to the optimal stopping time. Considering cross-validation stopping we answer the question: In what ratio the examples should be divided into training and testing sets in order toobtain the optimum performance. In the non-asymptotic region cross-validated early stopping always decreases the generalization error.Our large scale simulations done on a CM5 are in nice agreement with our analytical findings.