Alan Fortuny Sicart

FastAI for Deep Learning Deep learning is crashing it, there are many applications such as feature extraction, product recommendations, cancer detection, fire detection that can be done with deep learning. Without stepping into the wording appropiatness, in these post series, I share my journey on fastai. Fastai is a high level framework to train and deploy deeplearning models with little computation burden and very few lines of code. It is based on Pytorch which is the most popular DL framework as we speak. We are using deep learning for some years, but experimentation and deployment have taken far too long, and FastAI can be a great source to solve many problems. I particularly like the idea to have a framework that does not require a lot of computing and long learning curves. Let me be very honest, I struggle a lot with setting the environment and getting to run the code, but I persisted and I can only share how worth is it. You can find all the notebooks in the following repo: http

 When the images one is working with are of high resolution, or the demands of the application allow for very little margin of error  (autonomous driving, medical analysis) it makes sense to consider high resolution image translation. This post aims to summarize the following paper: 1711.11585.pdf (arxiv.org) High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs  Conditional GANs have been proven to provide fairly good translations from sketches to images, but its applications have been limited to lower resolution images. The following paper develops and algorithm that works on high resolution image translation 2048 × 1024 with very realistic generation. Introduction Creating realistic representations of the world is quite expensive computationally if every dimension and detail have to be model explictely. It becomes very necessary to find low weight approaches that could represent reality very realistic from a simple abstraction as an input. The following pap

The following post summarizes the following paper of the pix2pix model, a potential solution to create valid representations from sketches that are comparable from others that have images as input. 1611.07004.pdf (arxiv.org) Image-to-Image Translation with Conditional Adversarial Networks Conditional adverserial are a general purpose solution to image to image translation. It allows to reconstruct objects from sketches/edges maps, colorizing among many other tasks. The generic architecture of conditional adverserial NN seems to work well on a wide set of problems of image translation. Introduction The following method allows for the conversion of any input image type to the desired output image type, like we have seen with language models using the same architecture for multiple languages translation and other tasks. The main challenge with such image tasks is the definition of the loss function, which is not trivial to define for a given problem, while it has several consequences on t

 Our physical reality, and our products, have more and more a digital counterpart. The production lead times  of sports retail makes it challenging to have product samples when we need to encode visual product information for our advance analytics use cases, as we need to take decisions seasons before the product is in the store. One of the safest source on information that we have at this point in time is product sketch , a realistic abstraction of how the product will look like when produced . Textures and some details may be lost, but a lots of attributes such as colors, silhouette, design elements, patterns, technologies can be picked up by computer vision deep learning networks. It is therefore mandatory to be able to extract visual embeddings of the highest quality for this digital source. On our own data that improve coverage 77% for future seasons (around 14k products could have visual embeddings) In the following posts, I review what we understood as the most promissing appr

  A lot has been writen, since the book Limits to Growth in the 70' about the desirability to stop growth, as its costs exceed its benefits. There are wellbeing, natural and philosophical reasons to reach a steady state, to find what is enough, to celebrate and live with self impossed limits. The literature on the economic stability of such an economy is still rare, 50 years later, and Steffen Lange book, is the best contribution we have at this time. My last posts owe him almost everything I have to say about the topic. It was, the best selftaught course on keynessian and marxists economics I took, as it seems that in mainstream economics, there is over-representation of the neoclassical view of the economy, not leveraging what keynessian, marxists and ecological economics can offer. My prediction is that ecological economics will simply called economics, and the macroeconomics of the steady state will be called simply macroeconomics. In this post I will summarize what I have lear

 Overview It is a general theme that in a steady state economy, the aggregate supply and demand of goods become constant. That is no blocker for compositional (from dirty to clean), technological (from labor augmenting to resource augmenting) or enterprenourship change.  The achievemnt of a steady state economy is inevitable in the absence of technological progress. As long as there exists technological progress, the steady state could be postponed without sufficient interventions.  The tendency to invest more in labor saving technology, reduces the net demand from labor and requires goverment expenditures to keep artificially employment constant.  Working hours reductions are required to keep output constant in the steady state with labor augmenting technological progress such as AI. The wage increase should be equal to the hour reduction to keep purchasing power intact.  Constant demand requires contant investments to avoid ovecapacity, particularly to the same extent as the capital

The keynessian approaches to model the economy have focused on keep unemployment low, without barely paying attention to the environment. Despite this unfortunate fact, the keynessian mechanism to model the economy such as limited substitution of factors could help to analyze the economy consistently with main ecological economics principles . I will review only models with that limited substitution and skip the IS-LM-EE, despite being very popular among textbooks. Harris : Clean and Dirty Sectors Harris define the economy as a two sector economy, with a dirty-resource intensive and a clean-labor intensive sector. According to him there is a possible to make a composition change in the economy, which could lead to overall short term growth and long term steady state with lower environmental footprint.  That shift required f iscal and monetary policy to accelerate investments in the clean sector, which despite the overall lower productivity, could keep employment equal or even higher

In the previous post I cover the fundamentals of Keyness theory, without deeping dive in the goal of monetary policy in the steady state economy. This post aims to deep dive into the implications of monetary policy on growth and the conditions for a steady state economy. Davidson: Revenue Expectations and Monetary contrains Davidson transforms Keyne's theory into a long run theory. As finance plays a major role in the business cycle, it needs to be a part of a growth theory too. Firms are key decising the amount of investments in a given macroeconomic context. Due to the time lag between decision and production, the investors should make predictions of the evolution of prices, costs and demand. Investments increase capacity for multiple periods, while demand can fluctuate very rapidly, creating excesses or shortages. In order for the desired investment to happen, sufficient monetary capital should be available at affordable costs. The money supply is not given, as is rather a resul