Data Science Ryan (@DataScienceRyan) profile picture

This is @RyanAllred's #DataScience specific account. I run https://teapawt.net/t.co/hoJbq7nnMy where I teach Data Science.

Utah, USA
Joined March 2019
Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)Photo by Data Science Ryan (@DataScienceRyan)
RAISE YOUR RIGHT HAND AND REPEAT AFTER ME: We remove outliers when their extremeness indicates they're not part of our population — we do not get rid of them just because they are extreme!
3
9
5
109
19,468
Data Science Ryan retweeted
Just invested $65M into a new company that uses AI to slice bread. Welcome to the future:
438
123
83
3,590
747,034
Data Science Ryan retweeted
At Tonight's ML Workshop (8pm eastern) we'll be wrapping up our discussion of predictive modeling with Linear Regression by talking about Regularization techniques (Ridge, Lasso, ElasticNet) in addition to log transforming skewed x and y variables –resulting in our most powerful model yet! Join us at ->
1
1
1
300
Data Science Ryan retweeted
At tonight's Machine Learning Workshop (8pm Eastern) we'll be going over introductory basic best practices for Predictive Modeling with Linear Regression. Topics like: - Cross Validation - Filling NaNs - One Hot encoding - Feature Scaling - Pipelines
1
1
2
227
Data Science Ryan retweeted
Here's the recording of last week (Thursday's) Machine Learning Workshop where we learned about the basics of Gradient Descent. temzee.com/machine-learning-… We used Gradient Descent fit a simple linear regression model, and due to the model only having two parameters that enabled us to create some cool graphs and visualizations to better understand exactly how gradient descent discovers the model's best parameters. This week we'll be back to our regularly scheduled Tuesday at 8pm Eastern Machine Learning Workshop. Tomorrow we'll be fitting the absolute best predictive Linear Regression model that we can in the time that we have together. youtube.com/@temzee
1
1
176
Data Science Ryan retweeted
8
228
7
2,466
92,365
this is how you look when you comment your code
95
212
43
4,357
163,901
Data Science Ryan retweeted
Amjad just showed me Replit's latest AI code-generation stuff. When I was aghast at all the boilerplate it generated (probably no more than a normal Python app to be fair), Amjad said "Don't look at it. That's object code. You're programming in English now."
126
137
73
3,228
1,615,654
Data Science Ryan retweeted
A new paper in Nature found that you cannot, in fact, train AIs on AI-generated data and expect them to continue improving. What happens is actually that the model collapses and ends up producing nonsense.
619
2,140
821
15,495
1,877,297
Data Science Ryan retweeted
AI looks super sexy from the outside, but on the inside it’s about as unsexy as it gets. Endlessly tweaking and cleaning and piping tiny little pieces of data until you die.
17
12
4
198
49,908
Data Science Ryan retweeted
At tonight’s Machine Learning workshop (in 30 minutes at 8pm ET), We will be continuing our discussion of univariate analysis from last week with a focus on data visualizations. Livestream happening at -> youtube.com/@temzee We’ll talk about: - Data Visualizations for categorical vs continuous features - Histograms - Density Plots - Box Plots - Identifying outliers and what to do about them - Bar Charts - Stacked Bar Charts - Pie Charts We should be wrapping up univariate exploratory data analysis today, and next week we’ll be moving deeper into some introductory statistics topics as we move on to Bivariate Exploratory Data Analysis (which will end up being extremely important to our machine learning modeling). A reminder that I’ve added a new section to Temzee.com where you can browse all past Machine Learning Workshops and get access to any materials (notebooks, slides, etc) that were used during those sessions. You can also RSVP for future Workshops on this same page! -> temzee.com/machine-learning-…
1
1
105
Data Science Ryan retweeted
I've been working on writing Leetcode-ish Machine Learning challenges at temzee.com/code-challenges While I'm working on writing up a solid library of code challenges, I need some volunteers to help me beta-test the current format and to give me feedback on the quality of the prompts and the overall user experience so that I can make improvements. If you're into Python or Machine Learning, try your hand at a few of the introductory ones and let me know what you think of the format. ----------------------------------------- Temzee's code challenges are different from other sites because the in-browser notebook that I've built allows for the use of some basic Python libraries in addition to vanilla Python. So you *can* just use Python to complete the challenges, or if you want, you can sprinkle in some: - NumPy - Pandas - Scikit-learn - SciPy - Statsmodels - Regex etc. Because code challenges on Temzee allow for other Python modules, the number of possible working solutions is much greater than it would be if only vanilla Python were allowed. I think that this is a really good thing and will ultimately lead to users learning how to solve these challenges *both* with vanilla Python and *also* with tools that they're more likely to use on the day-to-day. It will naturally lead to so many important discussions about code style, efficiency, etc. Because of this, one of my top priorities is adding a section below each code challenge where users can submit and discuss potential solutions to the code challenges. Temzee's notebook-style code challenges can also be multi-step since I can break up the code challenge into multiple code cells and can grade the cells separately. Not only is this notebook-style more aligned with typical data science tooling, this opens the door for a greater variety of code challenges. Ultimately, if a person can't complete a challenge, I'd love for Temzee.com to be a top resource in teaching what they need to know to improve their coding skills. So as you can imagine there's a *looong* list of content that still needs to be created in order to enable that. If you have feedback you can message me here on X or on LinkedIn. Or, you can get ahold of me in the Temzee Discord Channel: discord.gg/wmjMEhz5jf I'm working on an easy way to report a bug or to leave feedback on specific code challenges directly on the code challenge page, but that probably won't be working until sometime next week.
1
4
170
Data Science Ryan retweeted
At tonight's ML Workshop (8pm Eastern at youtube.com/@temzee) We will begin our exploration of machine learning through the lens of applied mathematics, primarily Statistics and Linear Algebra. Tonight's discussion will focus on making sure that we have a good grasp of which math topics are important for different data professions. We'll discuss what the differences are between Data Analysts, Data Engineers, Data Scientists, Machine Learning Engineers, and Research Scientist roles, with a particular focus on how math preparation expectations differ between these roles. We'll take an in-depth look at Data Science roles specifically, and think about which math topics are the most useful and prevalent across data science. Then, together as a group, we'll decide on some applied-math-heavy data science projects to build together in future workshop sessions. If there are particular ML techniques or areas of math that you want us to focus on, make sure you tune in and vote in our polls: app.sli.do/event/gjhKbYJpzj6…
1
1
115
Data Science Ryan retweeted
Tonight's ML Workshop livestream is live! Join us to learn about data visualization and decision trees! youtube.com/watch?v=wL3luAL-…
1
1
105
Data Science Ryan retweeted
Tonight at 8pm Eastern we'll be continuing our discussion from Tuesday's stream on the Temzee YouTube channel: youtube.com/@temzee We'll explore the category of Supervised Learning in greater depth, we'll lay out a framework for how to approach machine learning modeling, we'll talk about the importance of starting with baseline models, and last (but not least) we'll peek under the hood of the DecisionTreeClassifier algorithm that we used at the end of Tuesday's livestream. See you then! If you want to RSVP for the stream you can sign up here: temzee.com/ml-workshop Or join the Discord Channel to become part of the Temzee Community: discord.com/invite/wmjMEhz5j…
1
3
100
Data Science Ryan retweeted
Every software company is scrambling for AI talent, while every AI company is scrambling to find one good backend engineer
11
58
16
780
72,400
Data Science Ryan retweeted
The massive AI productivity boost in software engineering compared to the pretty moderate/low AI productivity boost in many (most?) other industries means that software will just eat the world faster. – at San Francisco, CA
AI is going to change how software is created, but the world is still desperately going to need software engineers.
7
11
6
117
63,976
Data Science Ryan retweeted
A few months ago, Apple released their own Python-based ML framework with idea that users can make use of the extremely fast Apple Silicon hardware. It turns out that MLX is also really good for running LLMs, but depending on what kind of Mac you have, you may need to use a quantized LLM. In this upcoming #livecoding stream, we will demonstrate how we can use quantization to run an LLM on an average MacBook. See you this Friday (3/15) at 11:30am CT! #llm #genai #MLX
1
2
1
16
1,523