skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Rescuing neural spike train models from bad MLE
The standard approach to fitting an autoregressive spike train model is to maximize the likelihood for one-step prediction. This maximum likelihood estimation (MLE) often leads to models that perform poorly when generating samples recursively for more than one time step. Moreover, the generated spike trains can fail to capture important features of the data and even show diverging firing rates. To alleviate this, we propose to directly minimize the divergence between neural recorded and model generated spike trains using spike train kernels. We develop a method that stochastically optimizes the maximum mean discrepancy induced by the kernel. Experiments performed on both real and synthetic neural data validate the proposed approach, showing that it leads to well-behaving models. Using different combinations of spike train kernels, we show that we can control the trade-off between different features which is critical for dealing with model-mismatch.  more » « less
Award ID(s):
1734910 1845836
PAR ID:
10207528
Author(s) / Creator(s):
;
Date Published:
Journal Name:
Advances in Neural Information Processing Systems (NeurIPS)
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; (, Neural Computation)
    To understand neural activity, two broad categories of models exist: statistical and dynamical. While statistical models possess rigorous methods for parameter estimation and goodness-of-fit assessment, dynamical models provide mechanistic insight. In general, these two categories of models are separately applied; understanding the relationships between these modeling approaches remains an area of active research. In this letter, we examine this relationship using simulation. To do so, we first generate spike train data from a well-known dynamical model, the Izhikevich neuron, with a noisy input current. We then fit these spike train data with a statistical model (a generalized linear model, GLM, with multiplicative influences of past spiking). For different levels of noise, we show how the GLM captures both the deterministic features of the Izhikevich neuron and the variability driven by the noise. We conclude that the GLM captures essential features of the simulated spike trains, but for near-deterministic spike trains, goodness-of-fit analyses reveal that the model does not fit very well in a statistical sense; the essential random part of the GLM is not captured. 
    more » « less
  2. ; ; ; ; ; (, 2022 International Joint Conference on Neural Networks (IJCNN))
    Spike train decoding is considered one of the grand challenges in reverse-engineering neural control systems as well as in the development of neuromorphic controllers. This paper presents a novel relative-time kernel design that accounts for not only individual spike train patterns, but also the relative spike timing between neuron pairs in the population. The new relative-time-kernel-based spike train decoding method proposed in this paper allows us to map the spike trains of a population of neurons onto a lower-dimensional manifold, in which continuous-time trajectories live. The effectiveness of our novel approach is demonstrated by comparing it with existing kernel-based and rate-based decoders, including the traditional reproducing kernel Hilbert space framework. In this paper, we use the data collected in hawk moth flower tracking experiments to test the importance of relative spike timing information for neural control, and focus on the problem of uncovering the mapping from the spike trains of ten primary flight muscles to the resulting forces and torques on the moth body. We show that our new relative-time-kernel-based decoder improves the prediction of the resulting forces and torques by up to 52.1 %. Our proposed relative-time-kernel-based decoder may be used to reverse-engineer neural control systems more accurately by incorporating precise relative spike timing information in spike trains. 
    more » « less
  3. ; ; ; ; ; ; ; ; (, Oxford University Press)
    Pep-TCRNet is a novel approach to constructing a prediction model that can evaluate the probability of recognition between a TCR and a peptide amino acid sequence while combining inputs such as TCR sequences, HLA types, and VJ genes.Pep-TCRNet operates in two key steps:Feature Engineering: This step processes different types of variables:TCR and peptide amino acid sequencing data: The model incorporates neural network architectures inspired by language representation models and graph representation model to learn the meaningful embeddings.Categorical data: Specialized encoding techniques are used to ensure optimal feature representation for HLA types and VJ genes.Prediction Model: The second step involves training a prediction model to evaluate the likelihood of a TCR recognizing a specific peptide, based on the features generated in the first step. 
    more » « less
  4. ; (, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics)
    Neural models have become one of the most important approaches to dialog response generation. However, they still tend to generate the most common and generic responses in the corpus all the time. To address this problem, we designed an iterative training process and ensemble method based on boosting. We combined our method with different training and decoding paradigms as the base model, including mutual-information-based decoding and reward-augmented maximum likelihood learning. Empirical results show that our approach can significantly improve the diversity and relevance of the responses generated by all base models, backed by objective measurements and human evaluation. 
    more » « less
  5. ; ; (, International Conference on Pattern Recognition (ICPR))
    In this paper, we present a novel terrain classifica- tion framework for large-scale remote sensing images. A well- performing multi-scale superpixel tessellation based segmentation approach is employed to generate homogeneous and irregularly shaped regions, and a transfer learning technique is sequentially deployed to derive representative deep features by utilizing suc- cessful pre-trained convolutional neural network (CNN) models. This design is aimed to overcome the big problem of lacking available ground-truth data and to increase the generalization power of the multi-pixel descriptor. In the subsequent classification step, we train a fast and robust support vector machine (SVM) to assign the pixel-level labels. Its maximum-margin property can be easily combined with a graph Laplacian propagation approach. Moreover, we analyze the advantages of applying a feature selection technique to the deep CNN features which are extracted by transfer learning. In the experiments, we evaluate the whole framework based on different geographical types. Compared with other region-based classification methods, the results show that our framework can obtain state-of-the-art performance w.r.t. both classification accuracy and computational efficiency. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email