Source author record

Alexander Johnson

Alexander Johnson appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Artificial Intelligence Computation and Language eess.AS Genomics Neural and Evolutionary Computing Robotics Sound

Catalog footprint

What is connected

3works

7topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

A Reinforcement Learning-Based Model for Mapping and Goal-Directed Navigation Using Multiscale Place Fields

Autonomous navigation in complex and partially observable environments remains a central challenge in robotics. Several bio-inspired models of mapping and navigation based on place cells in the mammalian hippocampus have been proposed. This paper introduces a new robust model that employs parallel layers of place fields at multiple spatial scales, a replay-based reward mechanism, and dynamic scale fusion. Simulations show that the model improves path efficiency and accelerates learning compared to single-scale baselines, highlighting the value of multiscale spatial representations for adaptive robot navigation.

preprint2022arXiv

Automatic Dialect Density Estimation for African American English

In this paper, we explore automatic prediction of dialect density of the African American English (AAE) dialect, where dialect density is defined as the percentage of words in an utterance that contain characteristics of the non-standard dialect. We investigate several acoustic and language modeling features, including the commonly used X-vector representation and ComParE feature set, in addition to information extracted from ASR transcripts of the audio files and prosodic information. To address issues of limited labeled data, we use a weakly supervised model to project prosodic and X-vector features into low-dimensional task-relevant representations. An XGBoost model is then used to predict the speaker's dialect density from these features and show which are most significant during inference. We evaluate the utility of these features both alone and in combination for the given task. This work, which does not rely on hand-labeled transcripts, is performed on audio segments from the CORAAL database. We show a significant correlation between our predicted and ground truth dialect density measures for AAE speech in this database and propose this work as a tool for explaining and mitigating bias in speech technology.

preprint2014arXiv

Identifying the Genetic Basis of Functional Protein Evolution Using Reconstructed Ancestors

A central challenge in the study of protein evolution is the identification of historic amino acid sequence changes responsible for creating novel functions observed in present-day proteins. To address this problem, we developed a new method to identify and rank amino acid mutations in ancestral protein sequences according to their function-shifting potential. Our approach scans the changes between two reconstructed ancestral sequences in order to find (1) sites with sequence changes that significantly deviate from our model-based probabilistic expectations, (2) sites that demonstrate extreme changes in mutual information, and (3) sites with extreme gains or losses of information content. By taking the overlaps of these statistical signals, the method accurately identifies cryptic evolutionary patterns that are often not obvious when examining only the conservation of modern-day protein sequences. We validated this method with a training set of previously-discovered function-shifting mutations in three essential protein families in animals and fungi, whose evolutionary histories were the prior subject of systematic molecular biological investigation. Our method identified the known function-shifting mutations in the training set with a very low rate of false positive discovery. Further, our approach significantly outperformed other methods that use variability in evolutionary rates to detect functional loci. The accuracy of our approach indicates it could be a useful tool for generating specific testable hypotheses regarding the acquisition of new functions across a wide range of protein families.