Example 3

This example shows how to order topics according to co-usage in documents This will an ordering over topics where topics that co-appear often with other topics in documents to have similar indices.

This examples uses the OrderTopics function. In this procedure, for each topic, the probability distribution over documents is calculated. For each topic pair (i,j), the (symmetrized) KL distance between topic distributions i and j is calculated. The KL distances are then subjected multidimensional scaling (MDS). The first dimension of the MDS solution is then used to get an ordering for topics.

load a document-topic count matrix saved for the nips dataset

load 'ldasingle_nips';
load 'words_nips';

extract the topics in a cell array of strings

[S]=WriteTopics( WP,BETA,WO,5,0.6 );

order these topics using the OrderTopics function

[ Order ] = OrderTopics( DP,ALPHA );

and show the resulting ordering

S( Order )
ans = 

    'state policy optimal states function'
    'speech recognition system hmm training'
    'bayesian prior gaussian posterior distribution'
    'kernel support kernels risk svm'
    'theorem bound bounds proof dimension'
    'examples algorithm learning hypothesis set'
    'regression margin projection training set'
    'recognition character segmentation characters word'
    'generalization training error learning teacher'
    'call routing game traffic load'
    'variables variable conditional probability belief'
    'mixture likelihood em log density'
    'face detection false recognition faces'
    'information independent source sources separation'
    'words word context vowel stress'
    'sequence sequences language length symbol'
    'classification classifier training classifiers error'
    'user query text information queries'
    'node nodes tree trees decision'
    'clustering cluster clusters data algorithm'
    'class classes classification nearest neighbor'
    'training set test ensemble digit'
    'estimate estimation variance estimates regression'
    'training set data selection validation'
    'optimal complexity error criterion number'
    'basis rbf radial gaussian function'
    'prediction experts series expert predictions'
    'component components analysis principal pca'
    'algorithm algorithms convergence learning update'
    'probability distribution sample random samples'
    'vector vectors distance tangent euclidean'
    'parameter ai space method regularization'
    'matrix linear nonlinear vector matrices'
    'search block problem number blocks'
    'error gradient descent function backpropagation'
    'neural mlp transfer networks ann'
    'function functions linear approximation polynomial'
    'data number factor analysis variable'
    'information morgan advances systems kaufmann'
    'performance results set test number'
    'case section general assume defined'
    'learning learn learned'
    'task tasks architecture module modules'
    'layer training network hidden propagation'
    'net problem nets problems neural'
    'weight weights term decay small'
    'equation solution method equations set'
    'cost limit theory results simple'
    'energy optimization temperature annealing boltzmann'
    'state recurrent states transition order'
    'units unit hidden network connections'
    'network neural'
    'process approach stochastic simple type'
    'rules rule eeg knowledge neural'
    'noise gaussian noisy curve curves'
    'space map points dimensional mapping'
    'input output inputs outputs'
    'time path dynamic steps sequence'
    'activation constraints work university science'
    'level structure levels resolution hierarchical'
    'local global structure competitive locally'
    'model models modeling parameters'
    'rate figure time adaptation adaptive'
    'matching transformation invariant transformations rotation'
    'trajectory trajectories figure model point'
    'eq order case obtained phase'
    'patterns pattern input presented network'
    'representation representations connectionist distributed represented'
    'bit bits performance speed high'
    'field fields graph receptive graphs'
    'fig correlation correlations large analysis'
    'code population codes encoding reconstruction'
    'system behavior figure systems events'
    'parallel system neural processing elements'
    'signal time signals temporal processing'
    'control system controller systems forward'
    'activity subjects target position figure'
    'memory capacity associative stored memories'
    'image images feature features pixel'
    'frequency filter filters order low'
    'threshold gate binary boolean size'
    'system hand tracking video camera'
    'trials trial theory function press'
    'object visual objects recognition attention'
    'figure edge contour surface vision'
    'reinforcement action robot actions environment'
    'motion direction velocity moving flow'
    'dynamics system fixed state stable'
    'response stimulus stimuli responses contrast'
    'cells cell complex firing rat'
    'phase coupling oscillators oscillatory oscillator'
    'motor arm control movement hand'
    'neuron neurons spike firing information'
    'potential membrane current potentials voltage'
    'auditory sound channel localization spectral'
    'synaptic synapses excitatory inhibitory input'
    'activity neurons brain receptor sensory'
    'cortex orientation visual cortical connections'
    'eye head model gain map'
    'analog circuit chip current figure'