Unimodal Probability Distributions for Deep Ordinal Classification
نویسندگان
چکیده
Probability distributions produced by the crossentropy loss for ordinal classification problems can possess undesired properties. We propose a straightforward technique to constrain discrete ordinal probability distributions to be unimodal via the use of the Poisson and binomial probability distributions. We evaluate this approach in the context of deep learning on two large ordinal image datasets, obtaining promising results.
منابع مشابه
On a New Bimodal Normal Family
The unimodal distributions are frequently used in the theorical statistical studies. But in applied statistics, there are many situations in which the unimodal distributions can not be fitted to the data. For example, the distribution of the data outside the control zone in quality control or outlier observations in linear models and time series may require to be a bimodal. These situations, oc...
متن کاملThe unimodal model for the classification of ordinal data
Many real life problems require the classification of items into naturally ordered classes. These problems are traditionally handled by conventional methods intended for the classification of nominal classes where the order relation is ignored. This paper introduces a new machine learning paradigm intended for multi-class classification problems where the classes are ordered. The theoretical de...
متن کاملOn discrete a-unimodal and a-monotone distributions
Unimodality is one of the building structures of distributions that like skewness, kurtosis and symmetry is visible in the shape of a function. Comparing two different distributions, can be a very difficult task. But if both the distributions are of the same types, for example both are unimodal, for comparison we may just compare the modes, dispersions and skewness. So, the concept of unimodali...
متن کاملClassification and properties of acyclic discrete phase-type distributions based on geometric and shifted geometric distributions
Acyclic phase-type distributions form a versatile model, serving as approximations to many probability distributions in various circumstances. They exhibit special properties and characteristics that usually make their applications attractive. Compared to acyclic continuous phase-type (ACPH) distributions, acyclic discrete phase-type (ADPH) distributions and their subclasses (ADPH family) have ...
متن کاملA simple squared-error reformulation for ordinal classification
In this paper, we explore ordinal classification (in the context of deep neural networks) through a simple modification of the squared error loss which not only allows it to not only be sensitive to class ordering, but also allows the possibility of having a discrete probability distribution over the classes. Our formulation is based on the use of a softmax hidden layer, which has received rela...
متن کامل