Focus on These TOP 7 Data Science SKILLS in 2021

what's happening guys my name is nicholas and in this video we're going to take a look at the top seven skills you should be taking a look at in order to stand out as a data scientist machine learning engineer or deep learning engineer in 2021 ready to do it let's get to it skill number one skill number one focuses around the ability to use transformer architecture around computer vision and natural language processing tasks now you're probably thinking nick what the hell is a transformer are we talking about optimus prime or is it something completely different well you're exactly right it is something completely different so transformer architectures have been used to build large-scale natural language processing models for quite some time now but just recently what data scientists or researchers have actually found are they these same architectures are able to perform really really well on computer vision tasks so they've been able to perform well on super resolution on image detection object classification so they've been able to perform really really well now you're probably thinking why on earth should i focus on learning transformers well what's actually been found is that these same transformer architectures are able to outperform traditional convolutional neural networks they're able to scale significantly larger and perform more complex tasks now in order to get started with transformers i'd highly recommend you take a look at the tensorflow getting started tutorial but we'll be having a whole heap more of transformer content in the next coming months skill number two advanced natural language processing building word embeddings and performing text classification aren't new and advanced tasks that being said there's a whole heap of work being done on producing state-of-the-art natural language processing models organizations like open ai and hugging face are dropping heaps and heaps of research hours in order to build state-of-the-art models that allow us to do awesome things like generate blog posts from a short sample of posts other guys have actually used similar models to be able to generate entire plays from historical shakespearean documents so there's a whole heap of work being done in terms of the field of natural language processing hence why i think it's a really really important topic to get up to speed on and to get interested in in 2021 now in terms of getting started with nlp i highly recommend you check out my nlp playlist but also if you want to get started with some tensorflow basics that again there's a great bunch of tutorials there again all of the links that i mentioned in this video will be available in the description below skill number three skill number three focuses around generative adversarial neural networks also known as gans now gans are a neural network architecture that comprises of two key parts a generative component which actually tries to generate data this could be image audio picture video and a discriminator part which tries to detect whether or not that generated data is true or false or fake or not so the cool thing about gans is that they're being applied in a whole heap of really really interesting spaces things like taking low fps footage and increasing the number of frames per second but also they're being used in other interesting use cases like the ability to recolor video to generate artificial human faces and one of my favorites is its use in drug discovery so they're actually using gans to be able to generate new proteins to be able to treat certain illnesses now in order to get started with gans again we're going to be doing a lot of stuff on gans this year but if you want to get started by all means check out the tensorflow baseline tutorials there's a great simple gan tutorial that you can check out there skill number four this is probably my personal favorite skill number four is reinforcement learning reinforcement learning is a type of machine learning also known as semi-supervised learning where you actually train an agent inside of a simulated environment now the awesome thing about reinforcement learning is the fields that it's traditionally applied to and these three fields are probably some of my favorites so reinforcement learning is readily being used inside of game ai's so whenever you've seen videos on alphago or the deepmind team going out and building state-of-the-art models that are smashing chess records so on and so forth these have traditionally been reinforcement learning models now they're also being used in robotics as well as financial trading now the reason that i think that this is such an important skill to master is that it's still a new and emerging field but it will place such a huge amount of importance in the future being able to build or maintain models that operate in simulated environments is ultimately what's likely to get us to general ai so being able to work in that field is going to be highly beneficial if you're getting started inside of machine learning or deep learning now the best way to learn reinforcement learning is to check out some of the libraries that are already out there some of the ones that i've used already that i find really really good are keras rl tensorforce and rlib i've also done a couple of tutorials on this already link will be somewhere up there and also in the description below skill set number five skill set number five i think is one that's probably often overlooked but it's really really important skill set number five is being able to work as a full stack data scientist and integrate your models into different systems so lots of data scientists that i know tend to focus on building really really high quality models but often forget that these models need to be deployed back into embedded systems integrated into web apps or deployed into random environments so being able to have that skill set under your belt where you can not only build a model but also take that model all the way to deployment and integrate it back into a system is so so important and this is especially so given the fact that a lot of data scientists are now expected to take on more of a full stack role the best way to get started with more of a full stack data science skill set is to check out my full stack data science playlist so that'll be in the description below but also check out things like open ai and how to work with simulated environments as well as taking a look and broadening your skill set maybe look at how to develop web apps or work with raspberry pi's these skill sets are going to broaden your horizon as a data scientist and make you more valuable skill set number six skill set number six focuses around model explainability now as data scientists or ml engineers or deep learning engineers we tend to be focused on building awesome models to be able to solve interesting use cases that being said being able to explain how these models work how they generate predictions and opening up what we traditionally consider a black box is so important in terms of being able to take these models and finally deploy them into production this is why there's been such a focus on model explainability bias and fairness reduction as well as transaction transparency being able to explain how and why your models work makes it a whole lot easier to sell your idea and eventually get your project into production and integrated into the business environment now the best way to get started with model explainability is to play around with it in python so check out ai fairness 360 start interrogating your feature importance metrics from scikit learn and also interrogate your activation maps whenever you're building computer vision models with tensorflow that brings us to skill set seven skill set seven is one that i personally have been working a lot on towards the end of 2020 and is one that i'm definitely going to be working a lot on in 2021 and it revolves around the ability to read research papers and translate that research paper into code so i think the ability to read and understand state-of-the-art research papers goes a long way to understanding where the data science machine learning and deep learning industry is going and a lot of the new innovations that are coming out that being said being able to take the information the formulas the mathematical annotations and translate it into reproducible code i think is such a critical and important and valuable skill set that you can have in your arsenal this is particularly so because you can then take these state-of-the-art models and start to build them yourself this goes a long way in terms of demonstrating thought leadership and identifying you as a leader in the field now a great way to get started with this is to actually check out and start reading some of the new and emerging research papers and trying to start writing the code now some papers will come with code so you'll be able to use these as a guideline other times you'll have to work up to it that being said one of my absolute favorite websites to start getting started with this is actually papers with code papers with code has exactly what it says a lot of really interesting deep learning machine learning and data science research papers with their associated code this means you can go through the paper then go through the code and start to bridge the gap between those and that about wraps up our seven key skills thanks so much for tuning in guys hopefully you found this video useful if you did be sure to give it a thumbs up hit subscribe and tick that bell so you get notified of when i release future videos and let me know in the comments below if you think i left out any really important skills that you think are going to be important in 2021 thanks again for tuning in peace

Channel: Nicholas Renotte

Views: 6,518

Keywords: data science, data science 2021, data science skills, data science skills 2021

Id: sNNFEKZhOlA

Length: 9min 10sec (550 seconds)

Published: Sat Jan 02 2021