Data science/ML/AI
Kanalga Telegramāda oātish
Data science and machine learning hub Python, SQL, stats, ML, deep learning, projects, PDFs, roadmaps and AI resources. For beginners, data scientists and ML engineers š https://rebrand.ly/bigdatachannels DMCA: @disclosure_bds Contact: @mldatascientist
Ko'proq ko'rsatish
Tarmoq:š Telegram kanali Data science/ML/AI analitikasi
Data science/ML/AI (@datascience_bds) Ingliz til segmentidagi kanali faol ishtirokchi. Hozirda hamjamiyat 13 903 obunachidan iborat bo'lib, Texnologiyalar & Aralashmalar toifasida 8 919-o'rinni va Hindiston mintaqasida 29 117-o'rinni egallagan.
š Auditoriya koārsatkichlari va dinamika
Š½ŠµŠ²ŃŠ“Š¾Š¼Š¾ sanasidan buyon loyiha tez oāsib, 13 903 obunachiga ega boāldi.
26 Avgust, 2026 dagi oxirgi maālumotlarga koāra kanal barqaror faollikka ega. Oxirgi 30 kunda obunachilar soni 95 ga, soānggi 24 soatda esa -8 ga oāzgardi va umumiy qamrov yuqori darajada qolmoqda.
- Tasdiqlash holati: Tasdiqlanmagan
- Jalb etish (ER): Auditoriya oārtacha 8.25% darajada jalb etiladi. Nashrdan keyingi dastlabki 24 soatda kontent odatda umumiy obunachilar sonining 2.05% ini tashkil etuvchi reaksiyalarni toāplaydi.
- Post qamrovi: Har bir post oārtacha 1 146 marta koāriladi; birinchi sutkada odatda 285 ta koārish yigāiladi.
- Reaksiyalar va oāzaro taāsir: Auditoriya faol: har bir postga oārtacha 5 ta reaksiya keladi.
- Tematik yoānalishlar: Kontent panda, learning, row, api, ethic kabi asosiy mavzularga jamlangan.
š Tavsif va kontent siyosati
Muallif resursni shaxsiy fikrni ifoda etish maydoni sifatida taāriflaydi:
āData science and machine learning hub
Python, SQL, stats, ML, deep learning, projects, PDFs, roadmaps and AI resources.
For beginners, data scientists and ML engineers
š https://rebrand.ly/bigdatachannels
DMCA: @disclosure_bds
Contact: @mldatasci...ā
Yuqori yangilanish chastotasi (oxirgi maālumot 27 Avgust, 2026 da olingan) sababli kanal doimo dolzarb va katta qamrovli boālib qoladi. Analitika auditoriya kontent bilan faol hamkorlik qilishini, uni Texnologiyalar & Aralashmalar toifasidagi muhim taāsir nuqtasiga aylantirishini koārsatadi.
13 903
Obunachilar
-824 soatlar
-57 kunlar
+9530 kunlar
Postlar arxiv
13 903
Data Science Interview Questions and Answers.pdf13.55 MB
13 903
VC Dimension
In theory courses, VC dimension appears abstract.
But it answers a deep question:
How complex is your modelās decision boundary?VC dimension measures the largest number of points a model can shatter (perfectly classify in all labelings). Why this is importantā Two models with similar parameter counts can have very different capacities. For example: š¦ k-NN ā very high effective capacity š Linear classifier ā limited capacity š³ Deep trees ā extremely high capacity What you need to understand Generalization depends on capacity relative to data size. Too much capacity with little data leads to overfitting. ā VC dimension is about expressive power, not just number of parameters.
13 903
Repost from Programming Quiz Channel
Which ML concept refers to splitting data into training and testing subsets?
13 903
LLMs are getting insanely popular lately and suddenly everyone is talking about AI, chatbots, copilots, agentsā¦ so letās clear it up š
So what are LLMs really? š¤
LLMs = Large Language Models
Think of them as insanely smart text prediction machines that learned from tons of books, code, docs, and conversations šš»
Why everyone is obsessed right now š„
ā¢ They can write code š§āš»
ā¢ Explain complex stuff like a friend š£
ā¢ Analyze data š
ā¢ Power chatbots, copilots, agents š¤
ā¢ One model, MANY tasks
Why they exploded now š
ā¢ GPUs got better and cheaper
ā¢ Open source models became really good
ā¢ Companies realized: this saves time and money š°
The most famous LLMs you hear about š
ā¢ GPT-4 / GPT-4.1 by OpenAI
ā¢ Claude 3 by Anthropic
ā¢ Gemini by Google
ā¢ LLaMA 3 by Meta
ā¢ Mistral by Mistral AI
Where LLMs are actually used today š
ā¢ Chatbots and AI assistants
ā¢ Writing SQL and Python
ā¢ Data analysis and reporting
ā¢ Customer support automation
ā¢ Internal company tools
Important truth š”
LLMs are not magic šŖ
They are very powerful autocomplete with reasoning skills.
Learn how to use them properly and you are already ahead of most people š
13 903
š§ LayerNorm vs BatchNorm: Same Goal, Different Behavior
Both techniques normalize activations, but they operate differently.
Batch Normalization
š¦ Normalizes across the batch
ā”ļø Depends on batch statistics
š¼ Works very well in CNNs
ā ļø Sensitive to small batch sizes
Layer Normalization
š¬ Normalizes across features per sample
š Independent of batch size
š¤ Preferred in transformers and NLP
ā
Stable for sequence models
Why transformers use LayerNormā
Sequence models often run with variable or small batches.
LayerNorm avoids reliance on batch statistics and stays stable.
ā
Rule of thumb
š¼ CNNs ā BatchNorm
š¤ Transformers ā LayerNorm
š They look similar mathematically but normalize along different axes.
13 903
ā”ļøš One Line Feature Scaling
Scaling features without touching sklearn š
df["age_scaled"] = (df["age"] - df["age"].mean()) / df["age"].std()13 903
Python for Data Analytics: The Ultimate Library Ecosystem (2026 Edition)
This wheel is the Python data stack that's recommended from raw scraping to production insights:
ā”ļø Data Manipulation ā Pandas, Polars (the fast successor), NumPy
ā”ļø Visualization ā Matplotlib, Seaborn, Plotly (interactive dashboards)
ā”ļø Analysis ā SciPy, Statsmodels, Pingouin
ā”ļø Time Series ā Darts, Kats, Tsfresh, sktime
ā”ļø NLP ā NLTK, spaCy, TextBlob, transformers (BERT & friends)
ā”ļø Web Scraping ā BeautifulSoup, Scrapy, Selenium
š„ Pro tip from real projects:
šSwitch to Polars when Pandas starts choking on >1 GB datasets
š Use Plotly + Dash when stakeholders want interactive reports
š Combine Darts + Tsfresh for serious time-series feature engineering
13 903
š¤Æš Detect Outliers in 5 Lines
Simple Z score based outlier detection.
import numpy as np
z = (df["salary"] - df["salary"].mean()) / df["salary"].std()
outliers = df[np.abs(z) > 3]13 903
Pre-Chunking vs. Post-Chunking (On-Demand Chunking)
This visual breaks down two common ways to chunk documents in Retrieval-Augmented Generation (RAG) systems,and when each makes sense.
Pre-Chunking
Documents are cleaned, split into chunks, embedded, and stored ahead of time.
ā¢ Pros: Fast retrieval at query time, simpler runtime pipeline.
ā¢ Cons: Rigid,changing chunk size or strategy means reprocessing the entire dataset.
ā¢ Best for: Stable datasets, high-throughput apps, predictable queries.
Post-Chunking / On-Demand Chunking
Documents are stored whole; chunking happens after retrieval based on the userās query.
ā¢ Pros: More flexible and query-aware, often more relevant context.
ā¢ Cons: Higher latency and infrastructure complexity.
ā¢ Best for: Evolving content, exploratory queries, precision-focused use cases.
š Takeaway:
Thereās no one-size-fits-all. If speed and scale matter most, pre-chunk. If adaptability and relevance are key, post-chunk. Many production systems even combine both.
13 903
ā
Natural Language Processing (NLP) Basics You Should Know š§ š¬
Understanding NLP is key to working with language-based AI systems like chatbots, translators, and voice assistants.
1ļøā£ What is NLP?
NLP stands for Natural Language Processing. It enables machines to understand, interpret, and respond to human language.
2ļøā£ Key NLP Tasks:
- Text classification (spam detection, sentiment analysis)
- Named Entity Recognition (NER) (identifying names, places)
- Tokenization (splitting text into words/sentences)
- Part-of-speech tagging (noun, verb, etc.)
- Machine translation (English ā French)
- Text summarization
- Question answering
3ļøā£ Tokenization Example:
from nltk.tokenize import word_tokenize
text = "ChatGPT is awesome!"
tokens = word_tokenize(text)
print(tokens) # ['ChatGPT', 'is', 'awesome', '!']from textblob import TextBlob
TextBlob("I love AI!").sentiment # Sentiment(polarity=0.5, subjectivity=0.6)from nltk.corpus import stopwords
words = ["this", "is", "a", "test"]
filtered = [w for w in words if w not in stopwords.words("english")]
