Hey folks, today is a Mindblowing Monday 🤯!

Today I want to tell you about Language Models, a type of machine learning techniques that are behind most of the recent hype in natural language processing.

❓ Want to know more about them?

🧵👇 1/15

A language model is a computational representation of human language that models which sentences are more likely to appear in a given language.

🎩 Formally, a language model is a probability distribution over the sentences in a language.

❓ What are they used for?

👇 2/15

⚙️ Language models allow computers to understand and manipulate language at least to some degree. They are used in machine translation, speech to text, optical character recognition, text generation, and many more applications!

They come in many flavors:

👇 3/15

⭐ The simplest language model is the *unigram model*, also called a *bag of words* (BOW).

In BOW, each word is assigned a probability Pi, and the probability of a sentence is computed assuming all words are independent. But of course, this isn' true.

👇 4/15

For example, "water" is a more commonly used word than "philosophy", but the phrase "philosophy is the mother of science" is arguably much more likely than the phrase "water is the mother of science".

💡 The likelihood of a phrase depends upon all its words.

👇 5/15

THREAD on #TREND

How to Trade & Invest as per TREND ??

(1)

If TREND is BULLISH, BUY near key support levels & book profit near key resistance levels.
(Stop Loss below latest Higher Low (Swing Low) close basis on particular time frame)

#Equity4Life

👇

(2)

If TREND is BEARISH, SELL near key resistance levels & book profit near key support levels.
(SL above latest Lower High (Swing High) close basis on particular time frame)

Learn with #Equity4Life in real time market & with real traders.

#ZigZag #PriceAction #HH_HL #LL_LH

Try to learn from this tweet... It's helps to understand my trading & investing style. & Easily you understand my tweets & market views.

#Equity4Life #StockMarket
#Investing #Trading

Dataset bias strikes again! 😬

This is a great example of dataset bias! This ML method classifies skin lesions as being malignant or not.

Can you see the problem from the examples? 🤔

Let me shortly explain 👇

The dataset used to train the model contains images of both malignant and benign lesions. The problem is that the malignant images often contain rulers printed on the skin. They are used by doctors to assess the lesion over time.

Guess what the model learned... 👇

The ML model learned to detect the rulers... 🤷‍♂️

"In our dataset, images with rulers were more likely to be malignant; thus the algorithm inadvertently “learned” that rulers are

I have been working on 3 free, hands-on Python courses on the following fields:

1⃣ Algorithms
2⃣ Machine Learning
3⃣ Natural Language Procesing

Below are the content and links for each of them. I really hope they will be useful for you to learn more about these topics.

1/5

Content of the Algorithms course:

1. Algorithm's Basics
2. Divide and Conquer
3. Graphs
4. Dynamic Programming
5. Probabilistic Algorithms
6. Similarity Functions
7. Chaotic Systems
8. Cellular

Content of the ML course (work in progress):

1. Data science: basic techniques
2. Machine Learning: fundamentals, unsupervised learning and supervised learning
3. Deep learning: practical examples
4. Natural language processing:

Content of the NLP course:

1. NLP with spaCy
2. Semantic Enrichment of Entities: SPARQL and DBpedia
3. Spell Checker: from scratch and with PySpellChecker
4. Word Embedding with Gensim
5. Relationship between Words: d3.js and force

Any comments, questions or suggestions are welcome.

If you want to contribute to any of these free courses (add new content or translate it to another language), feel free to send a PR.

5/5

Beginners guide to start with machine learning.

What you need:

1. A programming language
2. A place to write and run code
3. A way to deal with data
4. A way to visualize results
5. A cool algorithm
6. An introductory course
7. A project to solve

Here are my recommendations.

Most people recommend Python to start.

I do too.

You can do machine learning with many different languages, but today, Python is the best option.

My advice: Get comfortable writing code before looking into machine learning.

You don't have to be a great developer to start.

But if you are, it helps.

Everything you know about building software is helpful to build machine learning systems.

Good developers have the odds in their favor.

"How much Python do I need to know before starting?"

If you ask, you aren't ready yet.

Ideal scenario: You should be comfortable building software. Most people get here within a year.

Let's move on to the 2nd item from the list: A place to write and run your code.

Get familiar with Google Colab.

• It's free.
• Requires no setup to start.
• It's available from anywhere.

Alternative: Kaggle.

What's machine learning?

Not a framework. Not a different language. Not the latest "cool" thing around the Internet.

Thread: Let me convince you that machine learning is the best thing since sliced bread 🍞.

Let's start with a simple exercise:

"Write a function that returns 0 if a supplied number is even or 1 if the number is odd."

If you are a developer, I'm sure you know how to write this function. It should be simple, right?

To avoid using a specific syntax, let's represent this function as follows:

y = f(X)

Where:
• X → The input number
• y → The result (0 if even, 1 otherwise)
• f → The function that does the work

We can now get to the interesting part.

If you are a programmer, writing this function is simple.

You know the relationship between X (the input number) and y (the output), so you can easily write the rules (the function.)

In machine learning, we don't write the rules. We learn them.

Imagine we had an algorithm that processes many samples (X → y), learns the relationship between them, and comes up with the correct rules.

We call these samples "training data," and they would like like this:

• 2 → 0
• 5 → 1
• 8 → 0
• 4 → 0
• 3 → 1
• ...

I ordered a pizza from @zomato recently. I thought this feature was a simple implementation, but this shocked me when I got to know the reality.

A Thread 🧵

We as a species hate to wait. With this feature, @zomato answered an impatient question by most people, 'When will my food arrive?'

Yes, I am talking about the Food Preparation time that @zomato displays on their app after the user orders food.

I thought that this is something that the restaurant enters. But @zomato uses Machine Learning to predict this time!

I was amazed after knowing this, and here is how they implement this to achieve accurate results.

Let's consider you also order pizza like me from 2 different restaurants. @dominos and a local restaurant that has the pizza on their menu but not their specialty.

It's easy to guess that @dominos will make the pizza faster as it's their specialty and will have a dedicated kitchen capacity geared towards preparing pizzas.

But more factors are in play here like queued orders, fine dining restaurants vs delivery kitchens, opening hours, etc.

One of the most common tasks performed in NLP is the identification of parts of speech (POS).

Therefore, it is useful to know the definition of the basic classes of POS:
1⃣ Noun
2⃣ Pronoun
3⃣ Verb
4⃣ Adjective
5⃣ Adverb
6⃣ Adposition
7⃣ Conjunction
8⃣ Interjection

1/4

1⃣ Noun (NOUN): is the name of a person, place, thing, or idea.

2⃣ Pronoun (PRON): is a word used in place of a noun.

3⃣ Verb (VERB): expresses action or being.

4⃣ Adjective (ADJ): modifies or describes a noun or pronoun.

2/4

5⃣ Adverb (ADV): modifies or describes a verb, an adjective, or another adverb.

6⃣ Adposition (ADP): is a word placed before (prepositions) or after (postpositions) a noun or pronoun to form a phrase modifying another word in the sentence.

3/4

7⃣ Conjunction (CCONJ or SCONJ): joins words, phrases, or clauses.

8⃣ Interjection (INTJ): is a word used to express emotion.

Finally, you can find the list of universal POS tags in the following link:

🔥 Applied Deep Learning - 2021 🔥

Just came across this absolute gem.

200+ free video lectures on all sorts of topics related to deep learning.

From Graph Attention Networks to RL algorithms.

repo: https://t.co/2IRF3CcR5p

playlist: https://t.co/N4zqkqnfdu

Insights from an open source influencer

I'm often asked how I get my content, over the years I've built an unusual technology stack for it

https://t.co/FdIAeZc8GF

Some insights:

I use Feedly for most content inputs because I can access the content through a single API endpoint and scraping is often pure pain.

@feedly

Feedly saves me a lot of time and manual

To pull and enrich my GitHub content I use ghapi from @fastdotai which provides a 100% always-updated coverage of the entire GitHub

The actual magic happens on my server infrastructure.

This is where the various data streams converge and are enriched.

The whole thing is based on Kafka streams so that if a pipeline stops it can be started up again without any problems.

Then a wild mix of machine learning NLP pipelines between @huggingface and @spacy_io is used to classify, score and tag the content.

So you think you know distillation; it's easy, right?

We thought so too with @XiaohuaZhai @__kolesnikov__ @_arohan_ and the amazing @royaleerieme and Larisa Markeeva.

Until we didn't. But now we do again. Hop on for a ride (+the best ever ResNet50?)

🧵👇https://t.co/3SlkXVZcG3

This is not a fancy novel method. It's plain old distillation.

But we investigate it thoroughly, for model compression, via the lens of *function matching*.

We highligh two crucial principles that are often missed: Consistency and Patience. Only both jointly give good results!

0. Intuition: Want the student to replicate _the whole function_ represented by the teacher, everywhere that we expect data in input space.

This is a much stronger view than the commonly used "teacher generates better/more informative labels for the data". See pic above.

1. Consistency: to achieve this, teacher and student need to see the same view (crop) of the image. For example, this means no pre-computed teacher logits! We can generate many more views via mixup.

Other approaches may look good early, but eventually fall behind consistency.

2. Patience: The function matching task is HARD! We need to train *a lot* longer than typical, and actually we were not able to reach saturation yet. Overfitting does not happen, as when function-matching, an "overfit" student is great! (Note: w/ pre-computed teacher, we overfit)

Checking for multicollinearity is a routine task while solving a data problem.

Most people rely on pairwise correlations for detecting multicollinearity which is a faulty approach in many scenarios

So what to do? Check this ↓

1/11

Multicollinearity creates a problem in the interpretation of the model when some predictors explain predictors.

Because then we are unable to understand the effect of each predictor in isolation towards the target variable and our coefficients become less useful.

2/11

Variance Inflation Factor or VIF is an efficient method to check for multicollinearity.

Variance Inflation refers to the inflation in the variance of the estimated coefficient of the independent variable because of the presence of multicollinearity.

3/11

We could have low pairwise correlations, but still have high VIF and vice-versa.

A strong relationship is possible between a predictor and other variables together combined, even though there's no high correlation individually.

4/11

Hence we prefer VIF

VIF performs a set of multivariate regression analyses to check the dependence among the independent variables by fitting multiple regression models on the dependent variables.

5/11

Time series: types and their nature

▪️Stochastic: the property of being well described by a random probability distribution.

▪️Deterministic: system in which no randomness is involved in the development of future states.

1/2

Therefore, stochastic time series can only be forecasted in a probabilistic sense.

Whereas a deterministic model will always produce the same result from a given starting condition or initial state.

2/2