Continual learning is a critical ability of continually acquiring and transferring knowledge without catastrophically forgetting previously learned knowledge. However, enabling continual learning for AI remains a long-standing challenge. In this work, we propose a novel method, Prototype Reminding, that efficiently embeds and recalls previously learnt knowledge to tackle catastrophic forgetting issue. In particular, we consider continual learning in classification tasks. For each classification task, our method learns a metric space containing a set of prototypes where embedding of the samples from the same class cluster around prototypes and class-representative prototypes are separated apart. To alleviate catastrophic forgetting, our method preserves the embedding function from the samples to the previous metric space, through our proposed prototype reminding from previous tasks. Specifically, the reminding process is implemented by replaying a small number of samples from previous tasks and correspondingly matching their embedding to their nearest class-representative prototypes. Compared with recent continual learning methods, our contributions are fourfold: first, our method achieves the best memory retention capability while adapting quickly to new tasks. Second, our method uses metric learning for classification, and does not require adding in new neurons given new object classes. Third, our method is more memory efficient since only class-representative prototypes need to be recalled. Fourth, our method suggests a promising solution for few-shot continual learning. Without tampering with the performance on initial tasks, our method learns novel concepts given a few training examples of each class in new tasks.

We establish the first nonasymptotic error bounds for Kaplan-Meier-based nearest neighbor and kernel survival probability estimators where feature vectors reside in metric spaces. Our bounds imply rates of strong consistency for these nonparametric estimators and, up to a log factor, match an existing lower bound for conditional CDF estimation. Our proof strategy also yields nonasymptotic guarantees for nearest neighbor and kernel variants of the Nelson-Aalen cumulative hazards estimator. We experimentally compare these methods on four datasets. We find that for the kernel survival estimator, a good choice of kernel is one learned using random survival forests.

We study online reinforcement learning for finite-horizon deterministic control systems with {\it arbitrary} state and action spaces. Suppose that the transition dynamics and reward function is unknown, but the state and action space is endowed with a metric that characterizes the proximity between different states and actions. We provide a surprisingly simple upper-confidence reinforcement learning algorithm that uses a function approximation oracle to estimate optimistic Q functions from experiences. We show that the regret of the algorithm after $K$ episodes is $O(HL(KH)^{\frac{d-1}{d}}) $ where $L$ is a smoothness parameter, and $d$ is the doubling dimension of the state-action space with respect to the given metric. We also establish a near-matching regret lower bound. The proposed method can be adapted to work for more structured transition systems, including the finite-state case and the case where value functions are linear combinations of features, where the method also achieve the optimal regret.

Model-free Reinforcement Learning (RL) algorithms such as Q-learning [Watkins, Dayan 92] have been widely used in practice and can achieve human level performance in applications such as video games [Mnih et al. 15]. Recently, equipped with the idea of optimism in the face of uncertainty, Q-learning algorithms [Jin, Allen-Zhu, Bubeck, Jordan 18] can be proven to be sample efficient for discrete tabular Markov Decision Processes (MDPs) which have finite number of states and actions. In this work, we present an efficient model-free Q-learning based algorithm in MDPs with a natural metric on the state-action space--hence extending efficient model-free Q-learning algorithms to continuous state-action space. Compared to previous model-based RL algorithms for metric spaces [Kakade, Kearns, Langford 03], our algorithm does not require access to a black-box planning oracle.

In this paper we address the problem of discovering a small set of frequent serial episodes from sequential data so as to adequately characterize or summarize the data. We discuss an algorithm based on the Minimum Description Length (MDL) principle and the algorithm is a slight modification of an earlier method, called CSC-2. We present a novel generative model for sequence data containing prominent pairs of serial episodes and, using this, provide some statistical justification for the algorithm. We believe this is the first instance of such a statistical justification for an MDL based algorithm for summarizing event sequence data. We then present a novel application of this data mining algorithm in text classification. By considering text documents as temporal sequences of words, the data mining algorithm can find a set of characteristic episodes for all the training data as a whole. The words that are part of these characteristic episodes could then be considered the only relevant words for the dictionary thus resulting in a considerably reduced feature vector dimension. We show, through simulation experiments using benchmark data sets, that the discovered frequent episodes can be used to achieve more than four-fold reduction in dictionary size without losing any classification accuracy.

We provide a comprehensive investigation of different custom and off-the-shelf architectures as well as different approaches to generating feature vectors for offensive language detection. We also show that these approaches work well on small and noisy datasets such as on the Offensive Language Identification Dataset (OLID), so it should be possible to use them for other applications.

This study's goal is to create a model of sentiment analysis on a 2000 rows IMDB movie comments and 3200 Twitter data by using machine learning and vector space techniques; positive or negative preliminary information about the text is to provide. In the study, a vector space was created in the KNIME Analytics platform, and a classification study was performed on this vector space by Decision Trees, Na\"ive Bayes and Support Vector Machines classification algorithms. The conclusions obtained were compared in terms of each algorithms. The classification results for IMDB movie comments are obtained as 94,00%, 73,20%, and 85,50% by Decision Tree, Naive Bayes and SVM algorithms. The classification results for Twitter data set are presented as 82,76%, 75,44% and 72,50% by Decision Tree, Naive Bayes SVM algorithms as well. It is seen that the best classification results presented in both data sets are which calculated by SVM algorithm.

Numeric representation of Text documents is challenging task in machine learning and there are different ways there to create the numerical features for texts such as vector representation using Bag of Words, Tf-IDF etc.I am not going in detail what are the advantages of one over the other or which is the best one to use in which case. There are lot of good reads available to explain this. It's a Model to create the word embeddings, where it takes input as a large corpus of text and produces a vector space typically of several hundred dimesions. The underlying assumption of Word2Vec is that two words sharing similar contexts also share a similar meaning and consequently a similar vector representation from the model. For instance: "Bank", "money" and "accounts" are often used in similar situations, with similar surrounding words like "dollar", "loan" or "credit", and according to Word2Vec they will therefore share a similar vector representation.

The persistence diagram is an increasingly useful tool arising from the field of Topological Data Analysis. However, using these diagrams in conjunction with machine learning techniques requires some mathematical finesse. The most success to date has come from finding methods for turning persistence diagrams into vectors in $\mathbb{R}^n$ in a way which preserves as much of the space of persistence diagrams as possible, commonly referred to as featurization. In this paper, we describe a mathematical framework for featurizing the persistence diagram space using template functions. These functions are general as they are only required to be continuous, have a compact support, and separate points. We discuss two example realizations of these functions: tent functions and Chybeyshev interpolating polynomials. Both of these functions are defined on a grid superposed on the birth-lifetime plane. We then combine the resulting features with machine learning algorithms to perform supervised classification and regression on several example data sets, including manifold data, shape data, and an embedded time series from a Rossler system. Our results show that the template function approach yields high accuracy rates that match and often exceed the results of existing methods for featurizing persistence diagrams. One counter-intuitive observation is that in most cases using interpolating polynomials, where each point contributes globally to the feature vector, yields significantly better results than using tent functions, where the contribution of each point is localized to its grid cell. Along the way, we also provide a complete characterization of compact sets in persistence diagram space endowed with the bottleneck distance.

We investigate the problem of classification in the presence of unknown class conditional label noise in which the labels observed by the learner have been corrupted with some unknown class dependent probability. In order to obtain finite sample rates, previous approaches to classification with unknown class conditional label noise have required that the regression function attains its extrema uniformly on sets of positive measure. We shall consider this problem in the setting of non-compact metric spaces, where the regression function need not attain its extrema. In this setting we determine the minimax optimal learning rates (up to logarithmic factors). The rate displays interesting threshold behaviour: When the regression function approaches its extrema at a sufficient rate, the optimal learning rates are of the same order as those obtained in the label-noise free setting. If the regression function approaches its extrema more gradually then classification performance necessarily degrades. In addition, we present an algorithm which attains these rates without prior knowledge of either the distributional parameters or the local density. This identifies for the first time a scenario in which finite sample rates are achievable in the label noise setting, but they differ from the optimal rates without label noise.