deltheil/vlfeat.torch

VLFeat (partial) FFI wrapper for Torch7

vlfeat.torch

VLFeat (partial) FFI wrapper for Torch7.

Runs with both LuaJIT and Lua 5.1 / 5.2 via luaffifb.

Install

luarocks make

You can run the tests with luajit -e "require('vlfeat.test')()".

Example

K-means clustering:

local vlfeat = require 'vlfeat'

local N   = 100
local D   = 64
local K   = 4

local kmeans = vlfeat.KMeans()

-- generate some random data (N x D-dimensional points)
local data = torch.rand(N, D)

-- run the algorithm with K centers (a.k.a. clusters)
kmeans:cluster(data, K)

-- get computed centers
local centers = kmeans:centers()
assert(
  (centers:size(1) == K) and
  (centers:size(2) == D)
)

-- pick some centers (for test purpose) and retrieve the corresponding indices
local points = centers:index(1, torch.LongTensor{1, 2, 4})
local assignments = kmeans:quantize(points)
assert(
  (assignments[1] == 1) and
  (assignments[2] == 2) and
  (assignments[3] == 4)
)