You gotta think about this one carefully!

Imagine you go to the doctor and get tested for a rare disease (only 1 in 10,000 people get it.)

The test is 99% effective in detecting both sick and healthy people.

Your test comes back positive.

Are you really sick? Explain below 👇

A surprising situation.

A machine learning model predicting 3 classes ends up with these results:

• Class A: 90% accuracy
• Class B: 80% accuracy
• Class C: 70% accuracy

Class C seems to be the worst-performing.

But you know already that I'll say it isn't. ↓

Here is the correct answer: We don't know yet which class is the worst.

This is important to understand, so keep reading.

Many people make this same mistake every day and try to fix things that ain't broken.

We are missing something, so let's go and get it.

How about if we ask a friend to look at the dataset manually?

Let's say these are the results of our friend doing the job:

• Class A: 95% accuracy
• Class B: 100% accuracy
• Class C: 71% accuracy

Do you see what just happened?

The model is actually really accurate when predicting Class C! 70% versus 71% when done by a human.

Class B, however, is a disaster: 80% versus 100% when done manually.

Class B seems to be the worse performing class!

The work that our friend did is called a "baseline."

Without having it, it's hard to determine how good are the results of a model.

I've seen tons of money down a drain just because teams didn't create a baseline beforehand.

But now, that won't happen to you 😉.

The 8-step quick-start guide to learn Machine Learning.

🧵👇

1⃣ Start with Python 🐍

Yes, you can do other languages, but Python is by far the most straightforward option.

👇

2⃣ Get familiar with numpy, pandas, and matplotlib

These three libraries are probably the most common Python libraries you'll have to use every day.

(Even if you don't end up doing machine learning, these libraries are awesome and useful.)

👇

3⃣ Start using notebooks

Look into Jupyter or Google Colab.

Notebooks are essential for data scientists and machine learning practitioners. Most of the code you'll read and write will be in notebooks.

👇

4⃣ Find a problem (already solved)

In my opinion, the best way to start is by working through a problem —especially when you can learn from its solution.

Start with something simple. I usually recommend "Titanic" from Kaggle.

👇

Machine learning education is broken.

If you are preparing for a research position, you are good. If you are looking to get out there and start solving problems, not even close.

Here are some thoughts so you can get ahead.

↓

Most classes, courses, and books cover the same road.

They start with a dataset. They finish with a working model. The focus is always on everything that happens in between.

Dataset → Model.

This is great, but not enough.

Real-life situations rarely start with a dataset, and they never end after you finish building your model.

Applying machine learning successfully is hard.

Here are a few examples that you should keep in mind.

First challenge: Properly framing up the problem.

If you don't understand the problem, you can't determine what data you need. If you don't understand the data, you can't build a good model.

I've never seen a company that had their data ready to go.

In fact, most of them don't even have data at all and need you to determine what exactly they should start collecting.

You usually have to go Problem → Potential Solution → Data.

One of the most popular activation functions used in deep learning models is ReLU.

I asked: "Is ReLU continuous and differentiable?"

Surprisingly, a lot of people were confused about this.

Let's break this down step by step: ↓

Let's start by defining ReLU:

f(x) = max(0, x)

In English: if x <= 0, the function will return 0. Otherwise, the function will return x.

If you draw this function, you'll get the attached chart.

Notice there are no discontinuities in the function.

This should be enough to answer half of the original question: the ReLU function is continuous.

Let's now think about the differentiable part.

A necessary condition for a function to be differentiable: it must be continuous.

ReLU is continuous. That's good, but not enough.

Its derivative should also exist for every individual point.

Here is where things get interesting.

We can compute the derivative of a function using the attached formula.

(I'm not going to explain where this is coming from; you can trust me on this one.)

We can use this formula to see whether ReLU is differentiable.

My recommendations to start with machine learning math:

• Statistics 110: Probability — Harvard University
• Essence of Linear Algebra — 3Blue1Brown
• Multivariate Calculus — Coursera

Links and 3 additional free resources here: https://t.co/fEmpmRKEvJ

If you are looking for more machine learning content, give me a follow and I’ll keep your feed interesting 😉.

“Should I give them more fluid?”

Extremely common question during AKI referrals from junior docs.

⛔️ Fluid management is tricky, often misunderstood, & can cause huge iatrogenic harm ⛔️

We try to convey important principles in THREAD 👇

#tipsfornewdocs #medtwitter #tweetorial

Currently too often patients with high creatinine are given too much IV fluid, driven by an AKI ➡️ fluid reflex

We’ve seen 7 litres be given in one shift.

Medics & surgeons, juniors & seniors, it seems everyone loves to reach for extra IV fluid (especially if urine output poor)

Why does huge volume IV fluid ‘resuscitation’ consistently occur?

Lots of reasons, including FALSE beliefs:
🛑 IV fluid always “treats AKI”
🛑 Hypotension = hypovolaemia
🛑 Lactate >2 means fluid deficiency
🛑 Septic patients are very hypovolaemic

The rule with IV fluids and AKI is to remember this 5 word phrase:

💥 💥 Aim for euvolaemia, then stop💥 💥

Fluid status is like Goldilocks - hypovolaemia is bad for the kidneys BUT excess volume is also harmful for renal function and will cause other complications too.

It is easy to say “target euvolaemia” - but we agree it is more difficult to do!

Fluid status assessment is certainly made harder by a belief that we must only rely on a snapshot assessment of examination features which are NOT specific / sensitive / reliable to elicit like.…

Want to include the menstrual cycle in your scientific work, but aren't sure how to do it? My lab (@katjaschmalen) have just published an article on best practices for observational studies of the cycle! Thread below! 🧵

Despite decades of research on the menstrual cycle, empirical studies have not adopted consistent methods for operationalizing the menstrual cycle, resulting in confusion in the literature and limited possibilities to conduct systematic reviews and meta-analyses.

Below, I summarize this "how to" article, and relay some of the key points!

PARTICIPANTS: All participants should be naturally-cycling people with ovaries (remember that they are not necessarily "women"! 🏳️‍⚧️ ⚧).

We provide a Reproductive Status Questionnaire with rules for identifying naturally-cycling people (current function of the sexual organs and hormonal/other medications that stop the cycle).

HELLO: If you pee and every time you finish, your urine looks like this picture, you may have a serious medical illness.

Foamy urine especially if it is persistent and noticeable over time may mean you have a serious kidney problem.

Let’s talk about this today...
RT for others.

What are the features of a normal urine?

A normal urine should have a light yellow colour, painless, odourless (or faint smell), steady stream and should ideally not foam when you pee.

When there’s something wrong, then you see changes in all the above normal features of urine.

For instance if urine becomes very deep yellow, it may mean you are dehydrated;

If urine becomes painful or smells really really bad, it may mean you have an infection;

If urine has a poor stream, especially in a man, it may mean a prostate problem like cancer.

Do you get me?

So today let’s talk about Foamy Urine.

What are the things that can cause urine to suddenly be foamy. As I said in the first tweet, it MAY mean a serious medical illness like a serious kidney problem.

First things first- it does NOT always mean a problem but it may as well.

Now what are the causes of Foamy Urine:

1. SPEED-
If you pee really fast or with a lot of force, it may make your urine more foamy than usual. The speed and force makes the urine generate a lot of bubbles which tend to almost quickly disappear as it is formed in the toilet bowl.

Here is a tweetorial from our latest publication in @Annals_Oncology about longitudinal tracking of esophageal adenocarcinoma. #OesophagealCancer #EsophagealCancer https://t.co/QWvrzBXmK7 1/8

We sequenced 245 plasma samples from 97 patients with oesophageal adenocarcinoma using a 77 gene pan-cancer ctDNA panel. 2/8

Variants derived from previously characterised driver oesophageal adenocarcinoma genes had a significantly higher VAF than variants from other genes, indicating selection. 3/8

Peripheral blood cell samples were also sequenced for 78/97 patients. CHIP mutations were identified in 23% of cases, longitudinal tracking of CHIP variants suggested these variants were dynamic over time. 4/8

We found patients that were ctDNA positive post-surgery had a significantly poorer survival than ctDNA negative patients, and the elimination of CHIP variants improved the positive predictive value. 5/8

I think that, in the face of an ongoing global health emergency and a pressing need to reopen safely, some form of health pass is a reasonable option to consider. But debate is a good thing, so here are some brief thoughts on various concerns I've heard about them... [thread]

"A health pass discriminates against people who have refused the vaccine." No. A health pass should also support proof of test result for people who haven't been vaccinated, and exemptions for people who cannot be vaccinated (get vaccinated if you can though, for heaven's sake).

"A health pass is an invasion of privacy." No. A health pass should be privacy preserving, allowing you to prove your status to third parties without giving away personal data or letting them track you without your permission.

"A health pass is unfair to people without smartphones." No. A health pass should accommodate people without a smartphone, e.g. with a printable option as a fall-back (though for most of us, carrying a digital pass on our phone will be much more convenient).

"A health pass will become obsolete as the virus mutates." No. A health pass should actually give us more insight and flexibility to adjust our response as our knowledge of the virus, and the ways we test / treat / vaccinate for it, increases.