Learning Future Classifiers without Additional Data

Kumagai, Atsutoshi; Iwata, Tomoharu

doi:10.1527/tjsai.d-h92

Cited by 5 publications

(8 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our work also closely relates to the field of Domain Adaptation, specifically the task of predicting future classifiers given past labeled data with timestamps [6], [8]. These are complementary to our work, as we use such solutions as a component of our framework.…”

Section: System and Demonstration Overviewmentioning

confidence: 94%

Just in Time: Personal Temporal Insights for Altering Model Decisions

Boer

Deutch

Frost

et al. 2019

2019 IEEE 35th International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

The interpretability of complex Machine Learning models is coming to be a critical social concern, as they are increasingly used in human-related decision-making processes such as resume filtering or loan applications. Individuals receiving an undesired classification are likely to call for an explanation-preferably one that specifies what they should do in order to alter that decision when they reapply in the future. Existing work focuses on a single ML model and a single point in time, whereas in practice, both models and data evolve over time: an explanation for an application rejection in 2018 may be irrelevant in 2019 since in the meantime both the model and the applicant's data can change. To this end, we propose a novel framework that provides users with insights and plans for changing their classification in particular future time points. The solution is based on combining state-of-the-art algorithms for (single) model explanations, ones for predicting future models, and databasestyle querying of the obtained explanations. We propose to demonstrate the usefulness of our solution in the context of loan applications, and interactively engage the audience in computing and viewing suggestions tailored for applicants based on their unique characteristic.

show abstract

Section: System and Demonstration Overviewmentioning

confidence: 94%

Just in Time: Personal Temporal Insights for Altering Model Decisions

Boer

Deutch

Frost

et al. 2019

2019 IEEE 35th International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

show abstract

“…The method cannot hence discover relationships between domains in an end-to-end fashion. Other methods like [7] and [8] consider only a logistic regression classifier and model the parameters as stochastic processes like vector autoregression (VAR) or a Gaussian process (GP). These methods require computing the posterior of GP-smoothed parameters, and this inference does not scale for multi-layered neural networks.…”

Section: Related Workmentioning

confidence: 99%

“…However, these domain adaptation approaches assume the presence of unlabeled data in the future, which is not available in our case. Another set of approaches aim to make network parameters a function of time [7,8,9] but they require external smoothness kernels as hyper-parameters for each parameter-type, which cannot be easily trained end-to-end on deep neural networks.…”

Section: Introductionmentioning

confidence: 99%

Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time

Nasery¹,

Thakur²,

Piratla³

et al. 2021

Preprint

View full text Add to dashboard Cite

In several real world applications, machine learning models are deployed to make predictions on data whose distribution changes gradually along time, leading to a drift between the train and test distributions. Such models are often re-trained on new data periodically, and they hence need to generalize to data not too far into the future. In this context, there is much prior work on enhancing temporal generalization, e.g. continuous transportation of past data, kernel smoothed time-sensitive parameters and more recently, adversarial learning of time-invariant features. However, these methods share several limitations, e.g, poor scalability, training instability, and dependence on unlabeled data from the future. Responding to the above limitations, we propose a simple method that starts with a model with time-sensitive parameters but regularizes its temporal complexity using a Gradient Interpolation (GI) loss. GI allows the decision boundary to change along time and can still prevent overfitting to the limited training time snapshots by allowing task-specific control over changes along time. We compare our method to existing baselines on multiple real-world datasets, which show that GI outperforms more complicated generative and adversarial approaches on the one hand, and simpler gradient regularization methods on the other.

show abstract

“…In contrast, the proposed method can infer domain-specific models by using the set of feature vectors in each domain. Some methods assume the additional semantic descriptors can be used that represent domains such as time information [19,17,18], device and location information [42], and instructions for reinforcement learning [28]. Although these semantic descriptors enhance the generalization ability, such data cannot be used under some circumstances.…”

Section: Related Workmentioning

confidence: 99%

“…It is important to infer an appropriate domain-specific model for each domain in the training phase since the characteristics of each domain differ. Existing methods infer such models by using additional semantic descriptors characterizing domains such as time information [19,17,18], device and location information [42], and pose and illumination information [29]. Although they improve performance, assuming semantic descriptors restricts the applicability of zero-shot domain adaptation because semantic descriptors cannot always be obtained.…”

Section: Introductionmentioning

confidence: 99%

Zero-shot Domain Adaptation without Domain Semantic Descriptors

Kumagai¹,

Iwata²

2018

Preprint

Self Cite

View full text Add to dashboard Cite

We propose a method to infer domain-specific models such as classifiers for unseen domains, from which no data are given in the training phase, without domain semantic descriptors. When training and test distributions are different, standard supervised learning methods perform poorly. Zero-shot domain adaptation attempts to alleviate this problem by inferring models that generalize well to unseen domains by using training data in multiple source domains. Existing methods use observed semantic descriptors characterizing domains such as time information to infer the domain-specific models for the unseen domains. However, it cannot always be assumed that such metadata can be used in real-world applications. The proposed method can infer appropriate domain-specific models without any semantic descriptors by introducing the concept of latent domain vectors, which are latent representations for the domains and are used for inferring the models. The latent domain vector for the unseen domain is inferred from the set of the feature vectors in the corresponding domain, which is given in the testing phase. The domain-specific models consist of two components: the first is for extracting a representation of a feature vector to be predicted, and the second is for inferring model parameters given the latent domain vector. The posterior distributions of the latent domain vectors and the domain-specific models are parametrized by neural networks, and are optimized by maximizing the variational lower bound using stochastic gradient descent. The effectiveness of the proposed method was demonstrated through experiments using one regression and two classification tasks.

show abstract

Learning Future Classifiers without Additional Data

Cited by 5 publications

References 22 publications

Just in Time: Personal Temporal Insights for Altering Model Decisions

Just in Time: Personal Temporal Insights for Altering Model Decisions

Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time

Zero-shot Domain Adaptation without Domain Semantic Descriptors

Contact Info

Product

Resources

About