[00:01:03] Welcome to another checkpoint episode. These checkpoints are basically delineations between levels of machine learning, learning. I. You can think of them as the finish line to individual machine learning campaigns. [00:01:23] So everything prior to this checkpoint is shallow learning fundamentals and starting now. Everything after this checkpoint is deep learning fundamentals. You can use these checkpoints as an opportunity to pause and go back and learn all of the details of shallow learning, for example, by way of the resources section. [00:01:43] Or you can forge on ahead into deep learning. Additionally, in these checkpoints, I'm going to boil down all of the resources that I've mentioned in prior episodes into the most fundamental resources. So if you want, you can ignore resources as I go through my episodes until these checkpoints and use them as sort of a reader's digest if you're pressed for time or something. [00:02:06] So let's talk about the most essential resources that I've mentioned in all the prior episodes. Resource number one, of course, being the Andrew ing Coursera course. Finish that. Get it over with. Move on by the way, I just found out, and I don't know how I missed this. Andrew ING has his machine learning course that he taught at Stanford University CS 2 29 recordings online, available for download on Stanford's website. [00:02:31] As well as iTunes U. It's basically the same thing as the Coursera course, but it actually covers significantly more algorithms and machine learning fundamentals. So I would recommend downloading these as audio supplementary material to support the Coursera course you're taking. I'll post that in the show notes in the audio section. [00:02:49] It's not audio, it's video MP four format, but you can just listen to the audio or convert it to MP three. From there, you'll want to learn Python, so you can actually start coding, find a good book or course or video series on the subject. After that, you're going to wanna learn psych, hit Learn, or TensorFlow. [00:03:05] Psych. Hit Learn is for shallow learning algorithms. TensorFlow is for deep learning algorithms. I recommend the TensorFlow website. The website's tutorials are as good as any book out there. At that point, if you want to go deeper into the shallow learning fundamentals, the core equations and algorithms, and where they come from, then read the Elements of Statistical Learning textbook. [00:03:29] As well as the pattern recognition and machine learning textbook, and after that you'll be prepared to start going down the rabbit hole of deep learning. Read the deep learning book.org book. I recommend doing these things in the order that I specified and maybe 45 minutes a day or a half hour a day, however much time you wanna spend learning machine learning. [00:03:51] Put aside a little bit of extra time for learning the math. Maybe 15 minutes a day or just one day a week, however you want to do it. Learn linear algebra, statistics, and calculus. All three of those from Khan Academy. And if you want good audio supplementary resources, I recommend the master algorithm and a handful of courses by the great courses that I've mentioned before. [00:04:14] One called Mathematical Decision Making and some courses on statistics and calculus. So per usual, all of these resources will be posted on my website, oc develop.com. That's OCD eve e l.com/podcasts/machine learning. If you've been around the internet trolling the world of machine learning, you may think that I'm missing some resources. [00:04:36] A very common one recommended is called artificial intelligence, a Modern Approach, a textbook about artificial intelligence, and another one is simply called Reinforcement Learning by Richard Sutton. Artificial intelligence and reinforcement learning are advanced topics. They come later, so I haven't missed these resources. [00:04:54] I'm simply holding off on them till later, everything in time. Also, in this episode, I want to give you an update on job hunting in machine learning. I did an episode on certificates and degrees, and since then I have interviewed like crazy trying to find a machine learning job. I'm lucky enough to have landed sort of a half and half position. [00:05:15] I'm doing half web development and half natural language processing. I will say that finding a job in machine learning is very difficult if you don't have a lot of background in this space. But what I have learned has confirmed my suspicions that I expressed to you in that prior episode. After many interviews, I found that indeed a PhD in machine learning. [00:05:35] Is unnecessary to find a job. It's necessary to find a job in research oriented positions such as those that deep mind and open ai. But if you just wanna be a machine learning engineer, you don't need a PhD. Even a master's is possibly not required. I at least landed all of my interviews with just a bachelor's of computer science, and nobody said a thing about lacking education. [00:05:54] The one thing that was my Achilles heel was lacking prior experience not having a solid portfolio of code to speak to. Just as I mentioned in the prior episode, you really want to build a strong portfolio, and I was rejected time and time and time again during these interviews. So now that you know the basics, what I recommend doing is start coding, do some weekend projects, pick up Python and TensorFlow, and start cracking away at that big idea you have in your head, the thing that's gonna make you a million dollars. [00:06:26] If you don't have an idea in your head of a project to work on, then the best recommendation, and I'm surprised that I didn't recommend this earlier, somebody pointed this out. There's a website called kagel, K-A-G-G-L e.com. It was acquired by Google recently and it is a website for participating in machine learning competitions. [00:06:45] So you join a team of people who are going to program some machine learning task. In a competition setting, what happens is a bunch of big companies like Netflix and Quora will put together some problem that they're actually trying to solve. Maybe whatever solutions they have internally aren't good enough and they need to take it a step up, so they'll release this problem on Kegel as a competition. [00:07:08] And any number of teams can participate in competing to solve this problem With machine learning, people will come up with different solutions, some shallow learning algorithms, some deep learning algorithms, and the winners of these competitions can get any number of prizes. There's swag, there's cash prizes, very large cash prizes, sometimes hundreds of thousands of dollars. [00:07:27] Another prize that you see on Kegel is employment opportunity. Sometimes the winners of these competitions just get hired by the company putting on the competition. But even if you don't get hired directly by a company and a competition, you can use competition participation as a portfolio piece when applying to companies. [00:07:46] And indeed companies are looking for that. They want see if you have any Kegel experience. I. So that's my update. Start building things. It's very important to start coding. Either work on a side project for some big idea you have, or start participating in Kegel competitions. That's it for this checkpoint episode. [00:08:05] Deep dive, those resources, and I'll see you next time.