Often with Natural Language Processing (NLP) applications a pipeline is useful to take the raw text and process it and extract relevant features before inputting it into a machine learning (ML) algorithm. Normalization From the standpoint of an ML algorithm, it may not make much sense to differentiate between different cases of a word – […]
In a previous post, I walked through the maths of back-propagation (“backprop”). Here I will go through the implementation in Python (heavily based on Andrew Ng’s course). I’m going to use the alternative form equations (in the last blog post I denoted those with a tilde, but now I will drop that tilde). This means […]
Jupyter Notebooks are a great and widely used tool in data science. Quite often then run are run on localhosts or have to be accessed via SSH tunnelling. This is not very convenient when you want to share the results presented in the notebook to members of your team who are non-technical. In this post, […]
I’m sure that every developer who has learned a little bit of ML dreams of applying this to the stock market and getting rich. At first glance RNNs and LSTMs seem perfect tools for the job. Something tells me it’s not going to be that easy, but it can be fun to try and this […]
Imputing missing values From the mean of a feature Country Name YEAR GDP ARUBA 1965 NULL ARUBA 1966 5.872478e+08 Say you have a dataframe for GDP by `Country Name` for each `year`, but some years are missing values. One way to deal with the missing values is to fill them in with the mean GDP […]
In simple Logistic Regression, we have the cost function $$\mathcal{L}(a, y) = -yln{(a)} – (1-y)ln{(1-a)}$$ where $a$ is the predicted value and $y$ is the ground-truth label on the training set (${0, 1}$). Why this function? The intuitive version is that if $y=1$ the second term goes away, and the first term that […]