Summary of slides Introduction In order to catch up with the progress of the field, I did a meta-survey, a survey of surveys, in different areas of deep learning. It allows for understanding the main challenges and trends in particular areas. Methodology I found the following papers by using the following keywords on Google, Google Scholar, Microsoft Academic, SemanticScholar, and others were found through connections on Connected Papers or other websites. I divide them into papers before 2020, and since 2020. The former includes 14 papers, while the latter includes 15, for a total of 29 papers. For all the papers I list the title, the date it was last updated (for most Arxiv papers), the number of citations (estimate of how impactful/useful it's been for others), the number of pages (how much content does it pack), and the first author (just to get familiar with some big names in the field). The classification in this section is relatively arbitrary, since the original purpose of this was to do a survey on GANs but then I expanded it to other topics. However, just for reference, the idea is that the blue DL/CV represented more general-purpose surveys, the green AS meant application-specific where it focused on a particular area or application, and the purple GAN represented general papers related to GANs.

What's PyTorch Lightning Fundamentally it's a wrapper for PyTorch training and testing code; it allows for encapsulation and organization of PyTorch models and corresponding associated components and functions, into Lightning Modules. PT Lightning Modules and PT Trainer: PT Lightning Modules help organize all the components of a machine learning system into one single class, including the training, validation and testing, along with optimizers, (and optionally data) into one single class. The PT Trainer takes the Lightning Module and helps abstract the loops and other mostly engineering aspects of the training and evaluation system, while keeping it easily accessible in case there needs to be any modification. Of particular noteworthiness it massively simplifies acceleration using CPU/GPU/TPUs and allows for easy parallel training, usage of mixed-precision, logging, and other (sometimes) cumbersome optimization tricks. Example: Pt Lightning Module For a simple image classification problem using lucidrains's implementation of the Vision Transformer (ViT) and an extended version of HuggingFace's configuration wrapper:

There's many online resources for learning computer vision. The goal of this post is to put some of (what I consider) the best, along with a plan and suggestions, to become an adept practitioner in this field. Motivation The last decade has seen as exponential increase in works that use computer vision (CV), and its endless derivative applications are only starting to get explored. The number of academic publications in the field is booming, and so is the interest from industries and governments, which is shown in an ever growing market all over the world. Computer vision research publications Computer vision market projection

Motivation If you're interested in starting research on anything related to artificial intelligence (AI), deep learning (DL) or its many applications, or like I was, a graduate student with not much experience reading papers, navigating through the relevant literature can feel like a daunting task. Every week there's probably dozens, if not hundreds, of new research articles posted to journals, conferences, and repositories like Arxiv. Keeping up with the literature is a challenge by itself, let alone getting started. This post aims to enable new researchers, like me, to familiarize themselves with the paper reading process. Reading a paper Someone once mentioned to me that some researchers intentionally over-complicate their papers to inflate the perception of how much work has been put into it, make it more difficult for others to (potentially) judge negatively, and make themselves appear smarter. I won't comment on this, but reading scientific papers, can definitely be challenging, stressing and time-consuming, specially at first. However, with time, and good technique, one gets used to the process.

Why learn how to program? Learning to write programs stretches your mind, and helps you think better, creates a way of thinking things that I think is helpful in all domains. -Bill Gates，Microsoft. I think everybody in this country should learn how to program a computer...because it teaches you how to think. -Steve Jobs, Apple Inc. In fifteen years we'll be teaching programming just like reading and writing...and wondering why we didn't do it sooner. -Mark Zuckerberg, Facebook.

This post is a collection of ideas, tips and tricks, along with supporting software and websites, for new graduate students (and anyone interested) to start doing research; It's a compilation of many scattered posts I've read throughout the years. Disclaimer: many of these are from the perspective of a CS student working on deep learning for computer vision, so your mileage may vary. Methodology summary The general methodology I would recommend to someone just starting on research is to: Find a rough topic or field you're interested on, and read an overview/review/survey article on the topic. This should allow you to get a general idea of the past and current state of the field, along with possible future directions. Obtain a more in-depth understanding of the field by going through the cited articles on the review. Summarize their contributions and identify trends in the literature, along with possible gaps in their works. Also, familiarize yourself with high-impact authors and where they publish, so as to stay up-to-date with their latest work. Find and solve a research problem. Maybe the most difficult part, but the general idea is that you either create a new problem (and a solution), or propose an improvement/solution to existing one.