Making sales for a brand new product can be challenging. You probably don’t have established processes or even a standard for what counts as “good” sales numbers. Despite that, you might be trying to hire salespeople, scale the entire business, or just win more of your deals.
Throughout years of working at startups and established businesses of many different sizes, I’ve had the opportunity to learn from some incredibly successful sales organizations. …
I love random forest models. They’re easy to set up, don’t require much power to train, and are easy to understand. It’s also one of the first models I used in production.
Random forest is an Interpretable Model, meaning you can see what it’s doing behind-the-scenes. Being interpretable is incredibly valuable when doing something like predicting which customers will cancel because instead of just knowing which customer will cancel, you can also know why.
What you probably didn’t know is that feature importance is calculated totally separately from the model training.
There are two popular ways to understand the “why”…
I’ve been writing Python for over 5 years and my toolset generally has been getting smaller, not bigger. A lot of tools just aren’t necessary or useful, and some of them you’ll simply outgrow.
These are three I’ve been stuck on for a long time, and unlike the rest, I just keep using them more and more.
Most code editors have an autocomplete feature that looks something like this:
This came across my feed today:
…and many gamers were outraged about it. I agree with most of them. The problem with this statement is the suggestion to “play at reasonable times”.
Playing at different times won’t help, to prove it we’re going to look at why playing online is just a small fraction of the bandwidth needed for gaming.
The majority of replies to that tweet were quoting this chart in defense of gaming:
These are games that support playing casually in a sandbox, where you don’t have to survive or win.
The great thing about games with a creative mode is the ability to get unlimited resources for building, without having to mine or gather.
As block-based games go, Boundless takes the cake in almost every way. There’s probably more to do and explore in this game than any other title in this list.
Planets upon planets of randomly generated landscapes combined with one universe of every player-built structure ever made.
Planets are connected through portals, and…
A great rule of thumb for writing code, especially in Python, is to look for a module on PyPi or just using Google, before you start writing code yourself.
If nobody else has done what you’re trying to do then you still might find articles, partial code, or general guidance.
If somebody has done it before you could find everything you need or at least examples of how others accomplished it or tried to.
In this case, generating fake data is something that many, many people have done before. A search for “fake data” on PyPi yields over 10,000 packages.
As someone who did over a decade of development before moving into Data Science, there’s a lot of mistakes I see data scientists make while using Pandas. The good news is these are really easy to avoid, and fixing them can also make your code more readable.
It’s nobody’s fault that there are way too many ways to get and set values in Pandas. In some situations, you have to find a value using only an index or find the index using only the value. …
Few things are certain when launching a business, but there is a consistent strategy that I’ve seen used in highly successful startups.
It took me a while to understand the nuances of this strategy, when it works and doesn’t work, and its many different forms. But over the years of founding, mentoring, and working inside of these companies, I’ve come to recognize four principles core to successful startups.
A friend of mine, Nate Watson, CEO of Contemporary Analysis, once told me that an early-stage startup only needs two things: sales and development. I’ve seen this proven over and over again.
It’s been almost two years now since my full-time job was at a company I founded. Since then I’ve gotten to work with some fantastic individuals that have changed some of my core beliefs about startups. Namely Nate Watson at Contemporary Analysis, Cory Scott at LiveBy, and the three founders of Buildertrend: Jeff Dugger, Steve Dugger, and Dan Houghton.
Learning from these people and many others has drastically changed my views on what’s important in an early-stage business. In hindsight, I think my mistakes as a founder were caused in part by core beliefs that were fundamentally wrong.
Unlike Neural Networks that use video card power (GPU), machine learning tasks with scikit-learn (sklearn) use processor power (CPU). Unfortunately, by default sklearn only uses one processor core, which means you’re probably only using a quarter of your CPU power!
Joblib is a Python module capable of many things, one of those is splitting tasks across your CPU. For this example, we’ll focus on training a model with sklearn.
Assuming you’ve already got your code written, you should have some lines near the end that look like this, where you train and test your model:
# Train the model
Startups, Business, Data Science, Product Management, and building stuff.