Inicio

José López-Collado

jlopez@colpos.mx

 Hi there!

In the dawn of the internet era (1995) I wrote a web site as a graduate student at Virginia Tech. It was a gratifying experience because I received e-mails from my country, Mexico and other places. At that time I use a text editor, a rudimentary knowledge of HTML and nightime hours, as usual for a graduate student. Now I am using this fancy automated scripting program from Google and launching a few pages that might be useful to the audience.

I am a professional entomologist and a Profesor  at Colegio de Postgraduados (COLPOS), Campus Veracruz. My job at COLPOS involves teaching and research. I taught two graduate courses:

I also manage these web sites (in spanish):

Introducción a la Estadística

Diaphorina citri y Huanglongbing en Mexico

My research is oriented to insect population ecology, risk analysis,  spatial analysis and simulation of biological populations. From time to time I am also conducting research in information systems and machine learning. In addition, I am exploring the opportunities to manage natural insect population to make craft work based on insect, specially tropical butterflies. This site is divided into pages that show some of my published work while at CP.


January, 2024

UMAP and species distribution modeling

We have published a paper in which we applied UMAP (Uniform Manifold Approximation and Projection) to select regions having a high similarity  with locations where a species occur:  Lopez-Collado, J. et al. 2024. Bioclimatic similarity between species locations and their environment revealed by dimensionality reduction analysis. Ecological Informatics, 79, 102444. https://doi.org/10.1016/j.ecoinf.2023.102444, 

The paper is Open Access:

https://www.sciencedirect.com/science/article/pii/S1574954123004739

https://doi.org/10.1016/j.ecoinf.2023.102444

UMAP is a dimensionality reduction technique, used to  analyze multivariate data. Usually, UMAP is applied to visualize the data in a reduced dimension (2D, 3D) to highlight patterns, clusters or separate classes. The purpose of the paper was to apply UMAP to measure similarity between background points and locations of a target species by analyzing bioclimatic variables associated to the given points. We use 10 pest species as examples.  Our hypothesis was that latent distance represented similarity in the multivariate, original space.  Therefore, close points are more similar than distant, separated points.  In this example, I am using Asbolus verrucosus as an example. 

Once we reduce the dimensions to 2 latent axes, we measure the intra-species distance, compute its empirical cumulative density function and then applied this function to compute the probability S of background-species distance.  The higher the probability, the smaller the distance. The next Figure shows in A the species-species distance (test to training points). In B the background-species distances are represented for training data.  Squares are species points and circles are background points.

The histogram below shows the species-species distance distribution and the plot on the right shows the equivalent probability of a background point near a species point. 

Once we compute the probabilities, we reattach the geographic coordinates to the tranformed data and visualize the map with the corresponding probabilities (next figure on left). The MESS analysis shows that the projection is within the range of the bioclimatic variables for the species (right plot).

The performance of the model is presented in the next table.

umap indices

We can see that the AUC is higher than 0.95, the standardized Kappa is 0.7, while the True Skill Statistics is also high ( > 0.92) and the binary correlation is positive. All of them indicate a relatively good model.


You can request the tutorial (in spanish), scripts and data to the author at: jlopez@colpos.mx and efialto@gmail.com

October 2017

Image Analysis Applied to Plant Disease Management (Papaya)

This year (2017) I am collaborating with Juan A. Villanueva-Jimenez and Eduardo Zozaya-Becerra (Cluster Institute) in a proyect to develop an automated  plant disease recognition system. This is an integrated project where specialists from the Universidad Autonoma de la Ciudad de Mexico: Rene Sagredo and Daniel Noriega are building the automated system. We at COLPOS are producing the papaya plants, studying the infectious process, taking pictures and making some image analysis to build a disease classifier, based on supervised learning. We are also planning to make multiespectral analysis with the appropriate mobile equipment.

This is a prototype built with Mathematica

Training set:

diseased leaves

healthy leaves

The classifier was trained using Logistic Regression

Applying the identifier to a diseased leave, p(diseased) = 0.99

Applying the identifier to a diseased fruit, p(diseased) = 0.99

Applying the identifier to a healthy leave, p(healthy) = 0.99

Last update: January, 2024