Jamshid Sourati

Classification active learning based on mutual information

Selecting a subset of samples to label from a large pool of unlabeled data points, such that a sufficiently accurate classifier is obtained using a reasonably small training set is a challenging, yet critical problem. Challenging, since solving this problem includes cumbersome combinatorial computations, and critical, due to the fact that labeling is an expensive and time-consuming task, hence we always aim to minimize the number of required labels. While information theoretical objectives, such as mutual information (MI) between the labels, have been successfully used in sequential querying, it is not straightforward to generalize these objectives to batch mode. This is because evaluation and optimization of functions which are trivial in individual querying settings become intractable for many objectives when we are to select multiple queries. In this paper, we develop a framework, where we propose efficient ways of evaluating and maximizing the MI between labels as an objective for batch mode active learning. Our proposed framework efficiently reduces the computational complexity from an order proportional to the batch size, when no approximation is applied, to the linear cost. The performance of this framework is evaluated using data sets from several fields showing that the proposed framework leads to efficient active learning for most of the data sets.

Accelerated Learning-Based Interactive Image Segmentation Using Pairwise Constraints

Algorithms for fully automatic segmentation of images are often not sufficiently generic with suitable accuracy, and fully manual segmentation is not practical in many settings. There is a need for semiautomatic algorithms, which are capable of interacting with the user and taking into account the collected feedback. Typically, such methods have simply incorporated user feedback directly. Here, we employ active learning of optimal queries to guide user interaction. Our work in this paper is based on constrained spectral clustering that iteratively incorporates user feedback by propagating it through the calculated affinities. The original framework does not scale well to large data sets, and hence is not straightforward to apply to interactive image segmentation. In order to address this issue, we adopt advanced numerical methods for eigen-decomposition implemented over a subsampling scheme. Our key innovation, however, is an active learning strategy that chooses pairwise queries to present to the user in order to increase the rate of learning from the feedback. Performance evaluation is carried out on the Berkeley segmentation and Graz-02 image data sets, confirming that convergence to high accuracy levels is realizable in relatively few iterations.

Active learning for efficient querying from a human oracle with noisy response in a language-model assisted brain computer interface

RSVP Keyboard™ is a non-invasive electroencephalography (EEG) based brain computer interface (BCI) for letter by letter typing. In this system a sequence of symbols is presented on a computer screen in rapid serial visual presentation scheme to query a users intent. EEG evidence and language model are used in conjunction for joint inference of the intended symbol. Usually repetition of sequences is necessary to achieve high confidence in the intended symbol selection. This repetition usually results in degradation in the speed of typing while compensating for accuracy. In this manuscript, we develop a mathematical framework for active sequence selection that would optimize the amount of evidence obtained from user and would improve both typing speed and accuracy simultaneously. Our analysis based on Monte-Carlo simulation shows that one can effectively improve both typing speed and accuracy by optimizing the sequence of queries to be asked from the BCI user.

A Probabilistic Active Learning Algorithm Based on Fisher Information Ratio

The task of labeling samples is demanding and expensive. Active learning aims to generate the smallest possible training data set that results in a classifier with high performance in the test phase. It usually consists of two steps of selecting a set of queries and requesting their labels. Among the suggested objectives to score the query sets, information theoretic measures have become very popular. Yet among them, those based on Fisher information (FI) have the advantage of considering the diversity among the queries and tractable computations. In this work, we provide a practical algorithm based on Fisher information ratio to obtain query distribution for a general framework where, in contrast to the previous FI-based querying methods, we make no assumptions over the test distribution. The empirical results on synthetic and real-world data sets indicate that this algorithm gives competitive results.

Automated localization of wrinkles and the dermo-epidermal junction in obliquely oriented reflectance confocal microscopic images of human skin

Reflectance Confocal Microscopic (RCM) imaging of obliquely-oriented optical sections, rather than with traditional z-stacks, shows depth information that more closely mimics the appearance of skin in orthogonal sections of histology. This approach may considerably reduce the amount of data that must be acquired and processed. However, as with z-stacks, purely visual detection of the dermal-epidermal junction (DEJ) in oblique images remains challenging. Here, we have extended our original algorithm for localization of DEJ in z-stacks to oblique images. In addition, we developed an algorithm for detecting wrinkles, which in addition to its intrinsic merit, gives useful information for DEJ detection.

An Active RBSE Framework to Generate Optimal Stimulus Sequences in a BCI for Spelling

A class of brain computer interfaces (BCIs) employs noninvasive recordings of electroencephalography (EEG) signals to enable users with severe speech and motor impairments to interact with their environment and social network. EEG based BCIs for typing popularly utilize event related potentials (ERPs) for inference. Presentation paradigm in current ERP-based letter by letter typing BCIs typically query the user with an arbitrary subset of characters. However, the typing accuracy and speed can potentially be enhanced with more informed subset selection. In this manuscript, we introduce the active recursive Bayesian state estimation (active-RBSE) framework for inference and sequence optimization. Prior to each iteration of presentation, rather than showing a subset of randomly selected characters, this framework optimally selects a subset based on a query function. Selected queries are made adaptively specialized for users during each intent detection. Through a simulation-based study, we assess the effect of active-RBSE on the performance of a language-model assisted typing BCI in terms of typing speed and accuracy. To provide a baseline for comparison, we also utilize standard presentation paradigms namely, row and column matrix presentation paradigm and also random rapid serial visual presentation paradigms. The results show that utilization of active-RBSE can enhance the online performance of the system, both in terms of typing accuracy and speed. Moreover, we conduct online experiments with human participants to study the human-in-the-loop effect on the performance of the proposed framework and the results were consistent with the simulations.

Assessing subsets of analytes in context of detecting laboratory errors

Laboratory error detection is a hard task yet plays an important role in efficient care of the patients. Quality controls are inadequate in detecting pre-analytic errors and are not frequent enough. Hence population- and patient-based detectors are developed. However, it is not clear what set of analytes leads to the most efficient error detectors. Here, we use three different scoring functions that can be used in detecting errors, to rank a set of analytes in terms of their strength in distinguishing erroneous measurements. We also observe that using evaluations of larger subsets of analytes in our analysis does not necessarily lead to a more accurate error detector. In our data set obtained from renal kidney disease inpatients, calcium, potassium, and sodium, emerged as the top-3 indicators of an erroneous measurement. Using the joint likelihood of these three analytes, we obtain an estimated AUC of 0.73 in error detection.

A novel delta check method for detecting laboratory errors

Investigating the variation of clinical measurements of patients over time is a common technique, known as delta check, for detecting laboratory errors. They are based on the expected biological variations and machine imprecision, where the latter varies for different concentrations of the analytes. Here, we present a novel delta check method in the form of composite thresholding, and provide its sufficient statistics by constructing the corresponding discriminant function, which enables us to use statistical and learning analysis tools. Using the scores obtained from such a discriminant function, we statistically study the performance of our algorithm on a labeled data set for the purpose of detecting lab errors.

Constrained spectral clustering for image segmentation

Constrained spectral clustering with affinity propagation in its original form is not practical for large scale problems like image segmentation. In this paper we employ novelty selection sub-sampling strategy, besides using efficient numerical eigen-decomposition methods to make this algorithm work efficiently for images. In addition, entropy-based active learning is also employed to select the queries posed to the user more wisely in an interactive image segmentation framework. We evaluate the algorithm on general and medical images to show that the segmentation results will improve using constrained clustering even if one works with a subset of pixels. Furthermore, this happens more efficiently when pixels to be labeled are selected actively.

Unsupervised wrinkle detection in reflectance confocal microscopy images of the human skin

Reflectance confocal microscopy (RCM) is a non-invasive and in-vivo imaging modality, which can take images from different depths of the human skin. A challenging problem is to detect a clinically important subsurface section of the skin, the Dermis/Epidermis junction, in RCM images. This is a tough problem because of the huge variation of texture and intensity features across both intersubject and intrasubject tissues. On the other hand, theres almost no wrinkle-free part of the skin. This well-known phenomenon can be used as a histological clue for guessing the probability of being Dermis or Epidermis in the neighboring regions. In this paper, we develop a two-step wrinkle detector for RCM images. By analyzing the results on different RCM images, we conclude it has high sensitivity and specificity, but a relatively lower Jaccard index.