Other articles

  1. Singular Valued Decomposition

    An alternative method for calculating a latent model relies on a mathematical miracle worker called the Singular Valued Decomposition (SVD) - what is this? Let's look at a graphical explanation. Here's what you get if you try and find the main parts of the “Lena” image (a famous image ...

    read more
  2. So you're Running a Data project

    A project workflow:

    Establish the Need - what is the actual value generated by the work? This stuff only works if you have some measurable outcome for the business. This helps with the often difficult sell to upper managament!

    Produce an "ideal" solution - Dream a little - how would it work if ...

    read more
  3. Machine Learning for the Common Man

    They say.....

    Supervised Learning

    1. Classification
    2. Regression

    You say....

    Educated Guesser given some known data

    1. Guesses words and Integer labels
    2. Guesses Real values

    They say....

    Unsupervised Learning

    1. Manifold Learning
    2. Clustering
    3. Latent Factor Analysis

    You say.....

    Explore relationships within the data

    1. Visually display relationships
    2. Group
    3. Find underlying drivers
    read more
  4. Feature Hashing in Coffeescript

    Feature Hashing is a useful technique for dealing with sparse or "bag-of-words" type data-sets. The essential idea is to hash the features into a new feature space of a prespecified vector size. The algorithm employs hashing on two levels - the first is the hashing to the hash-space index, the second ...

    read more

Page 1 / 3 »