https://jumpynitro.github.io/blog/ Recent content on Hugo -- gohugo.io en-us Thu, 07 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/mpcc/ Thu, 07 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/mpcc/ Generative adversarial model for clustering. This model is derived from a KL Divergence perspective between inference and generative distributions. https://jumpynitro.github.io/blog/2022/07/gim2/ Wed, 06 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/gim2/ Many Generative models have an infertence counter part. In this work we explore how they are formulated. We also studied their representation learning and generative capabilities theoretically and empirically. https://jumpynitro.github.io/blog/2022/07/gim1/ Tue, 05 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/gim1/ Many Generative models have an infertence counter part. In this work we explore how they are formulated. We also studied their representation learning and generative capabilities theoretically and empirically. https://jumpynitro.github.io/blog/2022/07/agp/ Mon, 04 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/agp/ Amortized Gaussian Process is a model that amortizes the computation of a Gaussian Process. With this model we can use autoencoder for time series with variable length and irregular sampling. https://jumpynitro.github.io/blog/categories123/ Sun, 03 Jul 2022 15:59:37 -0400 https://jumpynitro.github.io/blog/categories123/ fasdfdaafadfadfasdfsd https://jumpynitro.github.io/blog/about/ Sun, 03 Jul 2022 15:59:37 -0400 https://jumpynitro.github.io/blog/about/ My ML objective I am an ML researcher that is passionate, really passionate, about machine learning. I want to develop models capable of extracting the most information from all available data sources. At the same time, I want these models to generalize. In this development process I want to learn the most and help humanity using these autonomous models. Research interests I think that to develop such general models it is necesary to learn various ML subdisciplines: generative models, ML architectures, RL, transfer learning, causal structure learning, etc. https://jumpynitro.github.io/blog/research/ Sun, 03 Jul 2022 15:59:37 -0400 https://jumpynitro.github.io/blog/research/ There is a lot of current work from my thesis that I have not published yet. I currently working with Pablo Huijse, Pavlos Protopapas and Pablo Estévez to make this work available as papers. You can have an idea of this work here. The following work is already published. Publications Nicolas Astorga, Pablo Huijse, Pavlos Protopapas and Pablo A. Estévez. “MPCC: Matching Priors and Conditionals for clustering”, European Conference on Computer Vision (ECCV), Glasglow, 2020. https://jumpynitro.github.io/blog/industry/ Sun, 03 Jul 2022 15:59:37 -0400 https://jumpynitro.github.io/blog/industry/ Currently, I am working part-time in ALeRCE (Automatic Learning for the Rapid Classification of Events) as machine learning engineer. ALeRCE is a Chilean-led broker that process alert streams of astronomical data. The amount of astronomical data is huge and autonomous mechanisms are necessary to detect new interesting astronoimcal objects (note, we can not put a bunch of astronomers to look images of the sky). Being part of ALeRCE is my closest attempt to engineering, and it has been an entertaining journey. https://jumpynitro.github.io/blog/2022/07/lgad/ Sun, 03 Jul 2022 00:00:00 +0000 https://jumpynitro.github.io/blog/2022/07/lgad/ We present LGAd, a variational autoencoder for anomaly detection in time series. Anomaly detection is fundamental task in astroinformatics that could allows us to find new astronomical objects not observed in the past.