
Attempting to detect the number of important line segments in data
Many people using pwlf want to find out how many line segments are present in their data. There is an expensive way to do this, but this posts attempts to find some cheaper alternatives. A least squares fit and an Elastic Net are used to help identify important breakpoint locations.

Using the validate function to compare different FaceNet models in tindetheus
Comparing different facenet models on a hotornot dataset using the new validate function in tindetheus.

Battery and screen replacement on a LeEco Le Max 2 (x2)
I just replaced my battery and screen on my LeEco Le Max 2 and took pictures of the entire process.

Visual Studio Code Python Setup
Visual Studio Code is my favorite editor, and it's particular useful for working with Python. This post shows how to setup Code so that you can start writing better Python!

512 vs 128 FaceNet embeddings on Tinder dataset
Comparing the new 512 embedding facenet models with the old 128 embedding models on my Tinder dataset.

Force piecewise linear fit through data
Now you can use pwlf to force a fit through any set of data points! This is done as a constrained least squares problem.

pwlf update: fitting continuous piecewise linear models to data
New derivation used in pwlf library for fitting continuous piecewise linear functions to data.

Using facenet to automatically review Tinder profiles
I used facenet to find a pattern in the Tinder profiles I liked and disliked. This post explains how I did it.

Using tmpfs to speed up nonlinear finite element models
Benchmark of a HDD, SSD, and tmpfs (RAM disk) for read/write, and behncmark of an optimization on a nonlinear FE model.

Tindetheus: automatically like users on Tinder based on your own preference  How to install on Windows 10
Tindetheus makes it easy for a user to build and apply their own personalized machine learning model on Tinder. These models can be used to automatically like or dislike users, and take advantage of recent developments in computer vision to find a pattern in the faces you find attractive. The post goes on to describe how to install tindetheus on a Windows 10 PC.

Computers I've assembled
I've assembled a few computers throughout the years. Here is a list of the CPUs I've used and the year assembled for each of the computers I've put together.

Comparison of performance: Python NumPy and Numba, MATLAB, and Fortran
The time it takes to perform an array operation is compared in Python NumPy, Python NumPy with Numba accleration, MATLAB, and Fortran. The arrays are large, with one million to one billion elements. Numba is a great choice for parallel acceleration of Python and NumPy.

Comparing measures of similarity between curves
There are many different metrics that can be minimized to determine how similar two different curves are. This post looks at fitting a line to data points by minimizing different metrics of similarity. A least squares optimization is done which minimizes the sumofsquares. The traditional least squares fit is compared to minimizing the discrete Fréchet distance, the dynamic time warping (DTW) distance, and my own area metric.

Detect faces using facenet in Python
The TensorFlow implementation of facenet is an extremely powerful deep face recognition library. You can download pretrained models with a LFW accuracy of 0.993. This post demonstrates how you can use the facenet library to detect faces in images using Python.

Fitting a piecewise linear function to data
A Python library to fit continuous piecewise linear functions to data was created. This library allows for the user to specify the desired number of line segments when fitting piecewise linear functions.

Lack of Fit Test for Linear Regression
A lack of fit test is performed for simple linear regression model to see if the use of a linear model is appropriate for the given data set

Automating Tinder Likes with Support Vector Machine Learning
I recorded over 8,000 Tinder likes/dislikes so that I could use machine learning to predict whether I would like or dislike a new Tinder profile

Maximum Likelihood Estimation is Sensitive to Starting Points
Quick comparison of maximum likelihood estimation vs root mean square error for linear regression from random tarting points

Maximum Likelihood Estimation Linear Regression
Using maximum likelihood to fit a polynomial to data as opposed to a least squares fit

Gaussian Process Prediction (aka Kriging) with Different Correlation Functions
A simple one variable Kriging example utilizing scikitlearn to model a known function