CSE 250 B Assignment 4 Report
نویسندگان
چکیده
In this project, we implemented the recursive autoencoder (RAE) as described in Socher’s paper to discover the sentiment of sentences. We train and test our RAEs with a dataset of over 10000 sentences from movie reviews, and achieve 75.4% accuracy.
منابع مشابه
CSE 250 B Assignment 1 Report
In this report we analyzes two types of logistic regression models and carries out experiments using the UCI Adult dataset. We compare several factors that related to the prediction accuracy and discuss reasons of those correlations. Finally we propose a reasonable configuration of parameters and schemes for training models. Experiments show that for both models we achieved around 84% accuracy.
متن کاملCSE 250 B Assignment 3 Report
Latent Dirichlet Allocation (LDA) is a probabilistic, generative model designed to discover latent topics in text corpora, and it can be learned by collapsed Gibbs sampling. In this report, we evaluate the effectiveness of LDA by experiments on two dataset, Classic400 and BBC. We discuss related issues in Gibbs sampling, including goodness-of-fit criteria, parameter tuning, convergence, etc., a...
متن کاملAnodal Transcranial Pulsed Current Stimulation: The Effects of Pulse Duration on Corticospinal Excitability
The aim is to investigate the effects of pulse duration (PD) on the modulatory effects of transcranial pulsed current (tPCS) on corticospinal excitability (CSE). CSE of the dominant primary motor cortex (M1) of right first dorsal interosseous muscle was assessed by motor evoked potentials, before, immediately, 10, 20 and 30 minutes after application of five experimental conditions: 1) anodal tr...
متن کاملCoexisting causes of ischemic stroke.
BACKGROUND Coexistence of multiple potential causes of cerebral infarct (MPCI) has been poorly studied. OBJECTIVE To determine the risk factors, clinical findings, and topographical patterns of patients with at least 2 potential causes of cerebral infarct. DESIGN Data analysis from a prospective acute stroke registry (the Lausanne Stroke Registry, Lausanne, Switzerland) in a community-based...
متن کاملCse 250b Assignment 2 Report
In this report we describe our implementation of a conditional random field (CRF) model to divide novel English words into syllables. We use two different training models, namely Collin’s perceptron algorithm and stochastic gradient descent (SGD). The former provides word-level accuracy of 91.0%, and the latter 95.3%. Both are trained 10 epochs.
متن کامل