All Courses - Page 32 of 324 - Analytics Campus

Producing Extremely Efficient Decks for My Knowledge Science Class

Econometrics

January 30, 2026

Producing Extremely Efficient Decks for My Knowledge Science Class

Right this moment’s submit is in my sequence on utilizing Claude Code for the kind of analysis that utilized social scientists discover necessary, however this time it gained’t be analysis. Reasonably will probably be utilizing Claude Code to make your classroom lecture slides. I will likely be illustrating it “reside coding” so to see what’s doable. This lecture I’m targeted on is day 2 in an undergraduate knowledge science class at Harvard, within the authorities dept. They’re all nonetheless very early of their quantitative careers, and so I needed to give attention to 5 issues:

The rhetoric of quantitative social sciences.
Good empirical workflow (e.g., hierarchical folders, writing good code, automation, model management, replicability)

The video is lengthy at 1 hour and 40 minutes, however it’s like watching me paint a scene as a result of I actually don’t take the place that the optimum technique to do something in Claude Code is simply make some demand on the duty after which watch for it to be accomplished. I’d say I’m simply as intimately concerned within the job as I’d be with out AI. You be the decide for a way nicely I do it.

However the factor I attempt to emphasize on this video is that Claude Code is not only writing the decks for me utilizing latex as a result of it “is aware of methods to code in latex”. I feel that’s really the superficial understanding. It’s not merely doing that. Reasonably, it has as I’ve stated earlier than absorbed the tacit data concerned in what nice audio system who use decks to speak know however have in all probability by no means written down wherever, and should not even know methods to clarify it.

Dictating the Manufacturing of Your Lecture Slides

The way in which that I make my lecture slides utilizing Claude Code is thru exhaustive dictation. Discuss, in different phrases; share your concepts, your aims, the specified method; clarify your private beliefs about pedagogy; elaborate on what your objectives are, who your viewers is; share the massive image define in addition to the tiniest trivialities, because it all seems.

I name this dictation, not vibe coding. You’re creating the lecture by speaking into existence however you’re additionally modifying it always as you learn by means of the looks of every slide, in addition to get a really feel for the rhythm of the speak. So you will note me transfer issues round, tweak issues, strive new issues out, scrap outdated concepts solely, or remark them out.

What’s a terrific about dictation method to developing decks is that you could additionally introduce ideas that you simply solely vaguely take note of. As an example, I throw out a request the Claude Code undertake an method to what I’m recommend whereby the deck follows a rhythm such that the “marginal profit to marginal value ratio” throughout every slide is identical — a sort of “optimum rhetoric”. As an thought for what I would like, that’s each fairly correct and fairly far fetched on the similar. However in my thoughts, I feel I see how the right deck does typically comply with that sort of smoothness. I simply can’t say I possess the talent to do it, or the persistence, however I believe Claude Code can. And unsurprisingly, he does perceive what I imply and may make an try at it in how every slide connects to the following.

Unique theme and .sty file

One of many issues you will note on this video is the best way I requested it mid deck manufacturing to create a special beamer aesthetic design, and so it did — by actually creating from scratch its personal .sty model file!

And so when you watch it, you’ll see the model change the place I went from what felt like fairly boring grays and crimson colours to what I felt like was extra coloration, and extra fascinating kinds all through.

The opposite factor that I ask for are formidable drawings utilizing tikz, the highly effective graphics package deal in latex. Typically what Claude code can do seems like sorcery. Just like the submitting cupboard that he creates. Which I described anecdotally however didn’t explicitly request. I used to be speechless once I noticed it, and particularly after we perfected it collectively, and I’ve been doing this now for 2 months straight, since Thanksgiving actually. And I used to be nonetheless flabbergasted once I noticed it.

So so long as you might be prepared to stay immersed within the inventive and technical course of of manufacturing the deck on your lecture, I feel you will see it extremely helpful — a really actual reworking of the productiveness of your time, at the very least anyway for the modal tutorial. I feel the modal tutorial finds deck creation very tedious and you understand what my proof for that’s?

The truth that all tutorial beamer decks look kind of an identical to 1 one other, and that also they are normally dangerous, particularly mine. Not everyone seems to be born the Michelangelo of creating decks.

A Life Lived in Decks

I feel possibly 40% to 60% of my week entails being in a deck in a roundabout way, form or type. Both I’m making one or I’m utilizing one, however whichever it’s, decks are a nontrivial portion of my life.

Which implies I’m doing one thing with decks commonly. And they’re due to this fact each necessary on the facet of manufacturing and consumption. Getting them proper is due to this fact necessary and in addition very time consuming. Put one other means — time spent on decks has doubtless diminishing marginal returns, and rising marginal prices, and due to this fact we spend an period of time on that the place these two traces cross for every of us, and the standard of what we make is set by that high quality adjusted time use.

Properly that’s nonetheless true. However, I feel what’s modified is that the curves have shifted. Particularly, the marginal value curve has shifted left, and the marginal profit curve has shifted proper. And when that occurs, the period of time you spend on producing your deck is ambiguous — and you’ll see that performed out within the video as I spend an hour forty minutes and I’m not but carried out — however I’d contend that the standard of what you make goes up.

I feel the output to this point has been implausible. The decks have been high-quality tuned to me and my objectives, tailor made to the expertise of those particular college students I take note of. And you may simply search for your self right here at my instructing tab at both Gov 2001 (the PhD likelihood class) or 51 (the undergraduate knowledge science class) to see as I add the lectures in. And as soon as I attain a accomplished deck, I’ll additionally add the .tex file for these of who wish to see that too. Right here as an illustration is the lecture I did on foundational ideas in likelihood for the PhD college students.

Is Claude Code a Stage Shift, a Slope Change, or Each?

Sounds ridiculous to say this, however Claude Code has in all probability modified my life. I used to be attempting to clarify to somebody who makes use of ChatGPT intensively what it’s like. I principally stated “you understand how you felt like pre-ChatGPT you had been doing issues at x=100 however then after ChatGPT got here out, you felt such as you had been doing issues at x=1,000? That is principally like going from x=1,000 to x=1,000^2.

Share Scott’s Mixtape Substack

I do know that it feels like hyperbole to say that, however for me it isn’t hyperbole. It’s all solely attributable to the truth that Claude Code is within the folder with entry to the command line interface and due to this fact all these shell instructions. And I additionally needed to go away you with a couple of anecdotes too. First, listed below are two tweets (I bought this from Ethan Mollick’s LinkedIn). The one on the left is by somebody at OpenAI. The one on the suitable is by Boris Cherny, who created Claude Code. Have a look at what they each say when requested about how a lot of their code is now being produced by AI Brokers. Each of them, pc scientists thoughts you, say that 100% of the coding is now carried out at OpenAI and Claude by AI brokers. Not AI — not the copy-paste technique of taking the output from ChatGPT and placing right into a script. Not the autocorrect technique from Github copilot. No — brokers. Heck, Claude Code wrote 90% of its personal code itself!

Then take a look at this tweet that Andrej Karpathy, founding father of OpenAI, former director of AI at Tesla, stated. Apparently, me and him each began utilizing Claude Code two months in the past. I began I feel the center of November. I vaguely bear in mind utilizing it to do one thing, every week or so earlier than Thanksgiving, which elevated so much over thanksgiving once I had a challenge I wanted carried out, after which extra after that, after which massively throughout the Christmas break, then an enormous enhance in January the place I used to be spending 8-10 hours a day working feverishly prepping courses and ending initiatives that had stalled, to now the place it’s so deeply built-in into my means of working that I can’t even think about a world the place I may do with out it.

Andrej Karpathy@karpathy

A number of random notes from claude coding fairly a bit previous couple of weeks.

Coding workflow. Given the most recent raise in LLM coding functionality, like many others I quickly went from about 80% guide+autocomplete coding and 20% brokers in November to 80% agent coding and 20% edits+touchups in

8:25 PM · Jan 26, 2026 · 6.06M Views

1.49K Replies · 4.88K Reposts · 35.9K Likes

The fact is the educational curve in utilizing an AI Agent that will help you along with your courses, significantly writing decks, just isn’t solely flatter than folks assume it’s — it’s simpler. The educational curve could frankly even be negatively sloped for all anybody is aware of. It’s not simply straightforward to be taught; it isn’t even clear to me what you might be needing to be taught. There isn’t any magic “immediate engineering” talent anymore. That day the place you wanted to have some particular capability to speak is lengthy gone, if it ever did exist within the first place — it was actually all the time in all probability exaggerated by a bunch of snake oil salesmen on-line attempting to make a dime out of a nickel as “AI Influencers”.

There’s nothing to be taught greatest I can inform. There’s solely wanted a willingness to try to experiment. However that’s the way it frankly has all the time been with LLMs sensible worth to researchers and academics. Claude Code isn’t any completely different. It’s akin to some sort of magical sandbox the place you may principally dream up castles fabricated from gold and rubies utilizing sand, all utilizing solely your phrases. And its capability to do it? You give it the shell, and also you give it your folders.

We Are Not All Naturals, or Absolute vs Comparative Benefit

My former colleague, Rebecca Thornton, is a genius at many issues. She is difficult to maintain up with in principally any of them, not to mention be higher than her at any one in every of them. She is the basic instance of somebody who has absolutely the benefit in nearly each single a part of the duties of being a contemporary professor at a analysis college — nice instructor, nice researcher, and all of the duties concerned in every therein. And I’ve listened to her describe the method and imaginative and prescient of her courses earlier than — how a lot time she spends fascinated with the underlying structure of a course, what the coaching builds on prime of itself, methods of eliciting scholar engagement, how each factor reinforces each different factor resulting in human capital accumulation for the scholars and even herself over a 13 week interval — the period of time a semester lasts. It was past spectacular to look at, and that isn’t even referring to her analysis acumen. She merely is aware of methods to design a curriculum.

I’m not Rebecca Thornton a am I shut. She has absolutely the benefit in, greatest I can inform, each single job that accumulates linearly and nonlinearly within the creation of college class. And that completely consists of her slides. That are themselves based mostly on a rhetoric of her personal that she simply is aware of and is aware of nicely.

I frankly want Claude Code to simply shorten the hole between me and the professors. I probably can’t surpass them, however I can prolong my very own manufacturing chance frontier such that the expertise for college students in all probability goes up dramatically. My talent as a professor is stymied by my very own wrestle with inattention, obsession with the flawed issues, and misallocation of time and a focus throughout the myriad duties concerned in producing a curriculum and sophistication which incorporates however just isn’t restricted to the creation of decks.

I doubtless won’t ever have absolutely the benefit when put in a contest with the world’s greatest rhetorician, however I’m not attempting to both. I’m attempting to be my greatest self and Claude Code’s assist in writing decks is simply one of many methods I’m attempting to do this.

Optimum Manufacturing of Decks

I awakened early this morning considering of the issues I did within the video above final evening, and particularly how I used to be attempting to articulate this concept that there’s an optimum deck, and that it’s the place the marginal advantage of the time spent making the deck (which is probably going high quality diminishing in time and effort use) equals the marginal value. It is a deck I’m engaged on which I’m going to share now. Right here is the pdf (unfinished in the mean time), right here is the .tex and right here is the .sty file which is identical because the one which Claude Code invented on the fly for me within the video above once I requested him to modify as a result of I didn’t just like the aesthetics of what we had been utilizing at that second.

However, my rivalry has been that each curves shift with Claude Code particularly. Not LLMs typically — an AI Agent on par with Claude Code, particularly. One which lives within the listing. The impact that Claude Code has on the marginal profit curve is that for each unit of time spent making decks, the standard of these decks rise and therefore why the marginal profit curve shifts up. My instance is issues like creating aesthetically pleasing themes (which you’ll see me do within the video), accessing skilled stage talent at producing prime quality graphics with TikZ (a strong but totally bewildering LaTeX package deal for most individuals who worth their time at principally any non-zero worth), and the power to check out experimental concepts.

However the marginal value of manufacturing decks additionally doubtless declines. Right here I characterize that shift with a big decline, but additionally a flattening out, of the marginal value curve. It’s pointless to state simply what the brand new form is, however my private hunch is that it’s like this. The purpose the place we begin to actually hit the pressure on the underlying value perform with AI might be a lot farther to the suitable of even the hours in a day and as such my hunch is that the brand new marginal value curve is flatter which implies that at increased ranges of time spent working, the financial savings hole between the outdated marginal value and the marginal value grows and should even develop considerably at excessive ranges of use.

In order that’s an fascinating thought when you think about it. And this wraps it right into a easy provide and demand framework that I personally assume is so apparent to anybody utilizing Claude Code to make decks that it’s not solely noncontroversial to assert it, it’s borderline apparent. It’s clearly true to anybody who’s utilizing Claude Code to make their lecture decks that the time-adjusted high quality of your lecture slides goes up.

Being a Professor is a Assortment of Folders and Recordsdata

Claude Code for my part is for anybody whose job entails being in a listing of folders with recordsdata. That’s who that is for. It’s not only for writing R and Stata code. It could not even be primarily for that. It’s laborious to clarify, however it’s such as you’re in a position to manifest your goals by means of sheer needs. That’s in all probability a extra correct description of what I’ve been in a position to be taught to do with Claude Code than if I used to be to even attempt to say extra.

So, try the video. It’s lengthy, you may skip round. Share it. It’ll be free and never gated for an additional 2-3 days I feel. Every thing finally goes behind the paywall, so I encourage you to look at it instantly when you’re not a paying subscriber. However I feel seeing me make a deck for my precise class is value it and that alone in all probability will trigger a few of you to develop into instantly a subscriber. And bear in mind, I’m paying the $200-250/month or no matter it’s however that’s as a result of I’m an influence consumer. I’m utilizing this so intensively on a regular basis, as I’ve so many dormant initiatives on the verge of being orphaned that I’m lastly getting again up and operating. After all, provide creates its personal demand so I’m additionally doing new issues. However that’s okay. I have to and this has helped me to.

Thanks for studying Scott’s Mixtape Substack! This submit is public so be at liberty to share it.

Thanks once more on your help! Think about changing into a paying subscriber! It’s solely $5/month or $50/yr. You can too develop into a founding member which is $250 I feel and that’s principally simply you saying you wish to try this as I don’t have any add-ons. However no matter what you give, I may give you this.

Principled Coarse-Grained Acceptance for Speculative Decoding in Speech

Machine Learning

Dr. Mike

January 30, 2026

Principled Coarse-Grained Acceptance for Speculative Decoding in Speech

Speculative decoding accelerates autoregressive speech era by letting a quick draft mannequin suggest tokens {that a} bigger goal mannequin verifies. Nonetheless, for speech LLMs that generate acoustic tokens, actual token matching is overly restrictive: many discrete tokens are acoustically or semantically interchangeable, decreasing acceptance charges and limiting speedups. We introduce Principled Coarse-Graining (PCG), which verifies proposals on the degree of Acoustic Similarity Teams (ASGs) derived from the goal mannequin’s embedding area. By splitting every token’s likelihood mass throughout the overlapping teams that comprise it, we outline an overlap-aware coarse-grained distribution and carry out rejection sampling on the ensuing group variable. This yields an exactness assure on the group degree whereas permitting the accepted draft token to face in for any member of the group in apply. On LibriTTS, PCG will increase acceptance and throughput relative to plain speculative decoding and prior speech-specific relaxations whereas sustaining intelligibility and speaker similarity. These outcomes counsel acoustically conscious, group-level acceptance as a easy and normal strategy to speed up speech token era whereas sustaining speech high quality.

Apiiro’s Guardian Agent guards towards insecure AI code

Dr. Mike

January 30, 2026

Apiiro’s Guardian Agent guards towards insecure AI code

Apiiro has launched Guardian Agent, an AI agent that helps stop coding brokers from producing susceptible or non-compliant code by rewriting developer prompts into safe prompts, in keeping with the corporate.

Launched January 28, Guardian Agent is now in a non-public preview stage. Describing the expertise as introducing a basically new paradigm for securing software program within the period of AI-driven improvement, Apiiro stated Guardian replaces conventional appsec approaches constructed round detecting and fixing vulnerabilities after code is written. Guardian Agent replaces this reactive mannequin with a preventive one, stopping threat earlier than code is generated by guarding AI coding brokers in actual time, in keeping with Apiiro. Guardian Agent operates in actual time straight from the developer’s IDE and CLI instruments. The agent is powered by Apiiro’s code evaluation expertise and a software program graph that “deeply understands” the shopper’s software program structure and adapts to its adjustments, the corporate stated.

Elaborating on the inspiration behind Guardian Agent, Apiiro stated AI coding brokers are breaking the physics of software safety. Enterprises generate 4 instances extra code after adopting AI coding brokers and increase the appliance assault floor by six instances. This enlargement is pushed by speedy technology of latest APIs, duplicated open supply applied sciences and dependencies, and different assets, reshaping the software program structure with every code change, Apiiro stated. A lot of the code is generated with out builders being totally conscious of it. By stopping vulnerabilities earlier than code exists, safety outcomes are improved and developer productiveness is elevated, Apiiro careworn.

Polar bears are getting fatter within the fastest-warming place on Earth

Science

Dr. Mike

January 29, 2026

Polar bears are getting fatter within the fastest-warming place on Earth

Researchers tracked the physique situation of polar bears in Svalbard

Jon Aars, Norsk Polarinstitutt

Polar bears have been getting fatter whilst sea ice disappears in Norway’s Svalbard archipelago, the fastest-warming place on Earth – however scientists don’t anticipate the great occasions to final.

The northern Barents Sea, which stretches between Svalbard and Russia’s Novaya Zemlya within the Arctic Ocean, has been heating up seven occasions quicker than the globe as an entire. The ocean ice round Svalbard lasts two months much less in winter and spring than it did 20 years in the past. Bears now need to swim 200 to 300 kilometres between looking grounds on the ice and snow dens on the islands the place they provide delivery.

However the common measurement and weight of the Svalbard bears have elevated since 2000, a discovering that shocked Jon Aars on the Norwegian Polar Institute, who led the research.

“We must always take into consideration this as excellent news for Svalbard,” he says. “However if you’d like unhealthy information, you may simply go and look someplace else the place you’ve gotten very, very agency proof that local weather change is impacting polar bears negatively.”

This wide-ranging, solitary predator is cut up into 20 populations throughout the far north, the place this can be very troublesome to depend. Whereas its numbers are declining in components of Alaska, Canada and Greenland, they look like secure or growing somewhere else. For 9 of the populations, knowledge is just too sparse to say.

The Barents Sea inhabitants, which was estimated at 1900 to 3600 bears 20 years in the past, is regarded as secure or maybe even rising. Beginning in 1995, Aars and his colleagues tranquilised 770 bears with dart weapons from helicopters. They hopped out onto the snow or ice to measure their size and, to estimate weight, their girth on the chest.

Pattern evaluation confirmed this physique situation decreased till 2000, then elevated till the top of observations in 2019.

Polar bears rely on sea ice for a lot of facets of their lives

Trine Lise Sviggum Helgerud, Norsk Polarinstitutt

Within the spring, when ringed seals give delivery to pups on the ocean ice, polar bears hunt them to construct up shops of fats for the ice-free months. Aars and his colleagues imagine the shrinking ice space could also be making these seals simpler to search out.

The bears are additionally exploiting new meals sources. The roughly 250 people that stay on the islands when the ice recedes could also be looking extra bearded seals alongside the coast, in addition to harbour seals, that are spreading to Svalbard because the local weather warms.

These “native bears” are more and more ransacking duck and geese colonies for eggs, they usually have been seen chasing down reindeer from a rising cervid inhabitants. The carcasses of walruses, one other species that’s growing there, can present weeks of feasting.

Svalbard bears are higher in a position to adapt than scientists anticipated, “so extinction is delayed”, says Jouke Prop on the College of Groningen within the Netherlands.

“It’s a species in despair. They’re doing loopy issues,” he says. “It doesn’t work all over the place, however it might work for a while” on Svalbard.

Polar bears should not have reached the archipelago’s carrying capability after Arctic nations banned looking them for hides and zoo specimens in 1973. However warming is starting to disrupt the meals chain, which begins with algae on the underside of sea ice, warns Prop.

“It is going to be very troublesome to help an inexpensive inhabitants of polar bears if sea ice disappears,” he says.

“There might be a threshold, and… polar bears in Svalbard might be negatively affected by continued sea ice loss,” Aars says.

New Scientist. Science news and long reads from expert journalists, covering developments in science, technology, health and the environment on the website and the magazine.

The northern lights, fjords and glaciers: Svalbard and Tromso, Norway

Be part of an exhilarating Arctic journey in Norway, the place you may delve into the science behind the northern lights, Arctic ecosystems and human adaptation to excessive northern environments.

Subjects:

High 25+ Social Media Challenge concepts for College students

Statistics

Dr. Mike

January 29, 2026

High 25+ Social Media Challenge concepts for College students

Social media is not only for leisure. It has turn out to be an vital studying house the place college students can perceive communication, creativity, digital habits and know-how. Social media initiatives assist college students learn the way on-line platforms work, how content material spreads, and the way folks work together digitally. This weblog shares 25+ social media challenge concepts for college kids which are easy, sensible, and simple to grasp. Every challenge is designed to construct actual expertise like content material planning, evaluation, creativity, and accountable digital utilization. These initiatives are appropriate for varsity assignments, faculty initiatives, and inside assessments.

Additionally Learn: Social Media Analytics Developments: Key Insights & Future Prospects

Why Social Media Tasks Are Helpful for College students

Assist college students perceive digital communication.
Enhance creativity and content material planning expertise.
Train accountable on-line habits.
Construct presentation and analytical expertise.
Put together college students for contemporary careers.
Make studying interactive and sensible.

High 25+ Social Media Challenge concepts for College students

1. Social Media Consciousness Marketing campaign Challenge

Goal:
To unfold consciousness about an vital social difficulty utilizing social media.

Instruments Used:
Canva, Google Docs and primary analytics instruments

Anticipated Final result:
College students be taught content material planning, accountable posting and viewers engagement.

Platform Examples:
Instagram, Fb or X

2. Academic Instagram Web page Challenge

Goal:
To share academic content material in a easy and fascinating method.

Instruments Used:
Canva, Notion and smartphone digital camera

Anticipated Final result:
Improved clarification expertise and visible content material creation.

Platform Examples:
Instagram

3. Pretend Information Identification Challenge

Goal:
To take a look at how false info spreads on social media.

Instruments Used:
Browser analysis instruments, spreadsheets

Anticipated Final result:
College students discover ways to suppose critically and use the media.

Platform Examples:
Fb, X

4. Social Media Content material Calendar Challenge

Goal:
To plan content material upfront for constant posting.

Instruments Used:
Google Sheets, Trello

Anticipated Final result:
Understanding of scheduling, planning, and group.

Platform Examples:
Instagram, LinkedIn

5. On-line Polling and Survey Challenge

Goal:
To gather opinions utilizing social media polls.

Instruments Used:
Google Varieties, in-built ballot options

Anticipated Final result:
Knowledge assortment and primary evaluation expertise.

Platform Examples:
Instagram, X

6. Digital Storytelling Challenge

Goal:
To inform a narrative utilizing posts, captions and visuals.

Instruments Used:
Canva, smartphone

Anticipated Final result:
Higher storytelling and viewers connection.

Platform Examples:
Instagram, Fb

7. Psychological Well being Consciousness Web page

Goal:
To advertise constructive psychological well being messages.

Instruments Used:
Design instruments and analysis articles

Anticipated Final result:
Empathy growth and accountable communication.

Platform Examples:
Instagram

8. Social Media Advertising Fundamentals Challenge

Goal:
To grasp how manufacturers promote merchandise on-line.

Instruments Used:
Canva, analytics dashboards

Anticipated Final result:
Introduction to digital advertising and marketing ideas.

Platform Examples:
Instagram, Fb

9. Influencer Research Challenge

Goal:
To research how influencers construct belief and engagement.

Instruments Used:
Statement sheets, analytics instruments

Anticipated Final result:
Understanding of branding and viewers psychology.

Platform Examples:
Instagram, YouTube

10. Group Engagement Challenge

Goal:
To attach folks round a shared curiosity or trigger.

Instruments Used:
Posting instruments, remark monitoring

Anticipated Final result:
Improved interplay and moderation expertise.

Platform Examples:
Fb Teams

11. Social Media Ethics Challenge

Goal:
To check moral habits and on-line accountability.

Instruments Used:
Analysis sources, displays

Anticipated Final result:
Higher understanding of digital citizenship.

Platform Examples:
All main platforms

12. Picture Caption Writing Challenge

Goal:
To put in writing efficient captions for social media posts.

Instruments Used:
Picture editors, writing instruments

Anticipated Final result:
Improved writing and creativity.

Platform Examples:
Instagram

13. Quick Video Content material Technique Challenge

Goal:
To grasp how quick movies entice customers.

Instruments Used:
Cell digital camera, modifying apps

Anticipated Final result:
Video planning and modifying expertise.

Platform Examples:
Instagram Reels, YouTube Shorts

14. Social Media Analytics Challenge

Goal:
To trace likes, shares and engagement.

Instruments Used:
Constructed-in analytics, spreadsheets

Anticipated Final result:
Primary information interpretation expertise.

Platform Examples:
Instagram, Fb

15. On-line Security Consciousness Challenge

Goal:
To coach customers about on-line security.

Instruments Used:
Design instruments, analysis

Anticipated Final result:
Understanding privateness and safety points.

Platform Examples:
Instagram, Fb

16. Social Media Marketing campaign Comparability Challenge

Goal:
To match two completely different campaigns.

Instruments Used:
Case research, evaluation charts

Anticipated Final result:
Analytical and comparability expertise.

Platform Examples:
Instagram, X

17. Scholar Membership Promotion Challenge

Goal:
To advertise a pupil group or exercise.

Instruments Used:
Canva, scheduling instruments

Anticipated Final result:
Occasion promotion and teamwork expertise.

Platform Examples:
Instagram, Fb

18. Visible Branding Challenge

Goal:
To keep up constant colours and types.

Instruments Used:
Model kits, design instruments

Anticipated Final result:
Primary branding data.

Platform Examples:
Instagram

19. Remark Moderation Research

Goal:
To grasp how feedback have an effect on engagement.

Instruments Used:
Handbook monitoring, reviews

Anticipated Final result:
Battle dealing with and communication expertise.

Platform Examples:
Fb, Instagram

20. Hashtag Analysis Challenge

Goal:
To check how hashtags enhance attain.

Instruments Used:
Hashtag instruments, analytics

Anticipated Final result:
Search engine marketing-like understanding for social media.

Platform Examples:
Instagram, X

21. Social Media for Small Tasks Promotion

Goal:
To advertise a small concept or initiative.

Instruments Used:
Design instruments, posting schedules

Anticipated Final result:
Sensible promotion expertise.

Platform Examples:
Instagram, Fb

22. Social Media Pattern Evaluation Challenge

Goal:
To check trending subjects and challenges.

Instruments Used:
Pattern monitoring instruments

Anticipated Final result:
Understanding person habits.

Platform Examples:
Instagram, X

23. Content material Efficiency Comparability Challenge

Goal:
To match photos vs movies vs textual content posts.

Instruments Used:
Analytics, charts

Anticipated Final result:
Content material technique data.

Platform Examples:
Instagram, Fb

24. Social Media Language Research Challenge

Goal:
To check how language impacts engagement.

Instruments Used:
Publish samples, evaluation sheets

Anticipated Final result:
Communication and tone understanding.

Platform Examples:
All platforms

25. Person Engagement Enchancment Challenge

Goal:
To check methods to extend likes and feedback.

Instruments Used:
Analytics, experiments

Anticipated Final result:
Downside fixing and technique constructing.

Platform Examples:
Instagram

26. Social Media Time Administration Research (Bonus)

Goal:
To research every day utilization patterns.

Instruments Used:
Display screen-time reviews

Anticipated Final result:
Wholesome digital behavior consciousness.

Platform Examples:
All platforms

Find out how to Select the Proper Social Media Challenge

Choose a subject that pursuits you.
Select initiatives that contain studying, not copying.
Preserve goals easy and clear.
Give attention to clarification and understanding.
Use actual observations

How These Tasks Assist in Exams and Assignments

Social media initiatives assist college students perceive real-world digital habits as a substitute of studying concept solely. When college students work on initiatives, they achieve confidence in explaining ideas throughout displays and assessments. Academics usually worth originality, statement and readability, which these initiatives assist strongly. Sensible work additionally helps college students bear in mind ideas longer and reduces examination stress.

Conclusion

Social media initiatives provide college students a sensible approach to perceive how digital platforms work and the way folks talk on-line. These 25+ social media challenge concepts give attention to creativity, accountability and evaluation. By engaged on such initiatives, college students enhance communication expertise, digital consciousness, and important pondering. As an alternative of utilizing social media casually, college students be taught to watch, analyze and apply data meaningfully. With the appropriate challenge selection and energy, social media initiatives can turn out to be highly effective studying instruments that assist educational success and future alternatives.

Incessantly Requested Questions (FAQs)

1. What are social media initiatives for college kids?

Social media initiatives assist college students perceive how platforms are used for communication, advertising and marketing, consciousness, and content material creation. These initiatives give attention to planning, creativity, evaluation and accountable utilization moderately than simply posting content material.
2. Are these social media challenge concepts appropriate for inexperienced persons?

Sure, all challenge concepts are beginner-friendly. They’re designed in a easy method so college students can perceive ideas simply, even when they’ve primary data of social media platforms.
3. Which expertise can college students be taught from social media initiatives?

College students can enhance expertise comparable to content material planning, creativity, communication, information evaluation, teamwork, and digital accountability. These expertise are helpful for teachers in addition to future careers.
4. Can these initiatives be used for varsity or faculty assignments?

Completely. These initiatives are appropriate for varsity assignments, faculty initiatives, displays, and sensible assessments. Academics normally choose initiatives that present understanding and real-world utility.
5. Do college students want superior instruments to finish these initiatives?

No superior instruments are required. Most initiatives could be accomplished utilizing primary instruments like a smartphone, free design instruments, spreadsheets, and built-in analytics out there on social media platforms.

Unified Cisco Catalyst Middle Monitoring with CCGM APIs

Machine Learning

Dr. Mike

January 29, 2026

Unified Cisco Catalyst Middle Monitoring with CCGM APIs

Are you attending Cisco Stay Amsterdam and seeking to elevate your community monitoring expertise? Don’t miss the prospect to affix the DevNet Labs Studio session titled “Unified Cisco Catalyst Middle Monitoring: Constructing a Actual-Time Multi-Cluster Well being Report with CCGM APIs [DEVLAB-1100].” This hands-on workshop is designed to empower community engineers and builders alike with instruments to simplify and unify the administration of a number of Catalyst Facilities.

Simplify Multi-Cluster Administration with Cisco Catalyst Middle World Supervisor

Managing a number of Catalyst Facilities can typically really feel like juggling a number of advanced programs directly, every producing its personal telemetry knowledge and well being metrics. Cisco Catalyst Middle World Supervisor (CCGM) transforms this problem by providing a centralized, unified interface that aggregates knowledge throughout clusters. With CCGM, you achieve streamlined visibility and management, enabling you to observe community well being holistically somewhat than in remoted silos.

Construct a Actual-Time Cisco Catalyst Middle Monitoring Dashboard Utilizing CCGM APIs

What units this DevNet Labs Studio session aside is its sensible, hands-on method. Members will dive into the highly effective CCGM APIs to extract important telemetry knowledge from a number of controllers. Utilizing Python and HTML, you’ll construct a visually partaking, real-time multi-cluster well being report. This interactive dashboard will combination key efficiency metrics and current them in a way that’s each intuitive and actionable.

Not solely are you able to generate this report immediately, however additionally it is absolutely downloadable, making it simple to share insights together with your group or key stakeholders. This functionality is important for streamlining community operations and accelerating troubleshooting throughout distributed environments.

Why Attend This Cisco Stay DevNet Labs Session?

Study from the Consultants: Acquire direct entry to Cisco DevNet engineers and specialists who will information you thru the method.
Improve Your Expertise: Construct sensible expertise in API integration, Python programming, and real-time knowledge visualization tailor-made for community environments.
Enhance Operational Effectivity: Get palms on an actual CCGM dashboard and be taught the way it unifies monitoring, reduces complexity and improves decision-making.

Session Particulars

Ensure to order your spot in one of many following periods:

Monday, Feb 9 | 11:30 AM – 12:15 PM CET
Tuesday, Feb 10 | 4:00 PM – 4:45 PM CET
Wednesday, Feb 11 | 1:30 PM – 2:15 PM CET
Thursday, Feb 12 | 12:30 PM – 1:15 PM CET

You possibly can add this session on to your Cisco Stay schedule right here: Add DEVLAB-1100 to Your Schedule

Conclusion

Unified Cisco Catalyst Middle Monitoring Made Easy

For community professionals wanting to harness the total potential of Cisco Catalyst Facilities, this DevNet Labs Studio session is a must-attend. It presents an unparalleled alternative to grasp centralized community monitoring utilizing Cisco Catalyst Middle World Supervisor’s APIs and to create real-time, actionable well being experiences that may remodel the way you handle multi-cluster environments.

Don’t miss this opportunity to boost your community monitoring capabilities and join with fellow Cisco Stay attendees enthusiastic about innovation and operational excellence. See you at Cisco Stay Amsterdam!

State-of-the-art NLP fashions from R

Artificial Intelligence

Dr. Mike

January 29, 2026

Introduction

The Transformers repository from “Hugging Face” accommodates numerous prepared to make use of, state-of-the-art fashions, that are simple to obtain and fine-tune with Tensorflow & Keras.

For this goal the customers often must get:

The mannequin itself (e.g. Bert, Albert, RoBerta, GPT-2 and and many others.)
The tokenizer object
The weights of the mannequin

On this submit, we are going to work on a traditional binary classification process and prepare our dataset on 3 fashions:

Nevertheless, readers ought to know that one can work with transformers on quite a lot of down-stream duties, equivalent to:

characteristic extraction
sentiment evaluation
textual content classification
query answering
summarization
translation and many extra.

Conditions

Our first job is to put in the transformers bundle by way of reticulate.

reticulate::py_install('transformers', pip = TRUE)

Then, as traditional, load customary ‘Keras’, ‘TensorFlow’ >= 2.0 and a few traditional libraries from R.

Word that if operating TensorFlow on GPU one may specify the next parameters so as to keep away from reminiscence points.

physical_devices = tf$config$list_physical_devices('GPU')
tf$config$experimental$set_memory_growth(physical_devices[[1]],TRUE)

tf$keras$backend$set_floatx('float32')

Template

We already talked about that to coach a knowledge on the precise mannequin, customers ought to obtain the mannequin, its tokenizer object and weights. For instance, to get a RoBERTa mannequin one has to do the next:

# get Tokenizer
transformer$RobertaTokenizer$from_pretrained('roberta-base', do_lower_case=TRUE)

# get Mannequin with weights
transformer$TFRobertaModel$from_pretrained('roberta-base')

Information preparation

A dataset for binary classification is offered in text2vec bundle. Let’s load the dataset and take a pattern for quick mannequin coaching.

Break up our knowledge into 2 components:

idx_train = pattern.int(nrow(df)*0.8)

prepare = df[idx_train,]
check = df[!idx_train,]

Information enter for Keras

Till now, we’ve simply coated knowledge import and train-test cut up. To feed enter to the community we’ve to show our uncooked textual content into indices by way of the imported tokenizer. After which adapt the mannequin to do binary classification by including a dense layer with a single unit on the finish.

Nevertheless, we wish to prepare our knowledge for 3 fashions GPT-2, RoBERTa, and Electra. We have to write a loop for that.

Word: one mannequin generally requires 500-700 MB

# listing of three fashions
ai_m = listing(
  c('TFGPT2Model',       'GPT2Tokenizer',       'gpt2'),
   c('TFRobertaModel',    'RobertaTokenizer',    'roberta-base'),
   c('TFElectraModel',    'ElectraTokenizer',    'google/electra-small-generator')
)

# parameters
max_len = 50L
epochs = 2
batch_size = 10

# create an inventory for mannequin outcomes
gather_history = listing()

for (i in 1:size(ai_m)) {
  
  # tokenizer
  tokenizer = glue::glue("transformer${ai_m[[i]][2]}$from_pretrained('{ai_m[[i]][3]}',
                         do_lower_case=TRUE)") %>% 
    rlang::parse_expr() %>% eval()
  
  # mannequin
  model_ = glue::glue("transformer${ai_m[[i]][1]}$from_pretrained('{ai_m[[i]][3]}')") %>% 
    rlang::parse_expr() %>% eval()
  
  # inputs
  textual content = listing()
  # outputs
  label = listing()
  
  data_prep = perform(knowledge) {
    for (i in 1:nrow(knowledge)) {
      
      txt = tokenizer$encode(knowledge[['comment_text']][i],max_length = max_len, 
                             truncation=T) %>% 
        t() %>% 
        as.matrix() %>% listing()
      lbl = knowledge[['target']][i] %>% t()
      
      textual content = textual content %>% append(txt)
      label = label %>% append(lbl)
    }
    listing(do.name(plyr::rbind.fill.matrix,textual content), do.name(plyr::rbind.fill.matrix,label))
  }
  
  train_ = data_prep(prepare)
  test_ = data_prep(check)
  
  # slice dataset
  tf_train = tensor_slices_dataset(listing(train_[[1]],train_[[2]])) %>% 
    dataset_batch(batch_size = batch_size, drop_remainder = TRUE) %>% 
    dataset_shuffle(128) %>% dataset_repeat(epochs) %>% 
    dataset_prefetch(tf$knowledge$experimental$AUTOTUNE)
  
  tf_test = tensor_slices_dataset(listing(test_[[1]],test_[[2]])) %>% 
    dataset_batch(batch_size = batch_size)
  
  # create an enter layer
  enter = layer_input(form=c(max_len), dtype='int32')
  hidden_mean = tf$reduce_mean(model_(enter)[[1]], axis=1L) %>% 
    layer_dense(64,activation = 'relu')
  # create an output layer for binary classification
  output = hidden_mean %>% layer_dense(models=1, activation='sigmoid')
  mannequin = keras_model(inputs=enter, outputs = output)
  
  # compile with AUC rating
  mannequin %>% compile(optimizer= tf$keras$optimizers$Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0),
                    loss = tf$losses$BinaryCrossentropy(from_logits=F),
                    metrics = tf$metrics$AUC())
  
  print(glue::glue('{ai_m[[i]][1]}'))
  # prepare the mannequin
  historical past = mannequin %>% keras::match(tf_train, epochs=epochs, #steps_per_epoch=len/batch_size,
                validation_data=tf_test)
  gather_history[[i]]<- historical past
  names(gather_history)[i] = ai_m[[i]][1]
}

Reproduce in a Pocket book

Extract outcomes to see the benchmarks:

Each the RoBERTa and Electra fashions present some further enhancements after 2 epochs of coaching, which can’t be mentioned of GPT-2. On this case, it’s clear that it may be sufficient to coach a state-of-the-art mannequin even for a single epoch.

Conclusion

On this submit, we confirmed the right way to use state-of-the-art NLP fashions from R.
To know the right way to apply them to extra complicated duties, it’s extremely advisable to overview the transformers tutorial.

We encourage readers to check out these fashions and share their outcomes under within the feedback part!

Corrections

If you happen to see errors or wish to counsel adjustments, please create a difficulty on the supply repository.

Reuse

Textual content and figures are licensed beneath Inventive Commons Attribution CC BY 4.0. Supply code is accessible at https://github.com/henry090/transformers, except in any other case famous. The figures which were reused from different sources do not fall beneath this license and will be acknowledged by a observe of their caption: “Determine from …”.

Quotation

For attribution, please cite this work as

Abdullayev (2020, July 30). Posit AI Weblog: State-of-the-art NLP fashions from R. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2020-07-30-state-of-the-art-nlp-models-from-r/

BibTeX quotation

@misc{abdullayev2020state-of-the-art,
  creator = {Abdullayev, Turgut},
  title = {Posit AI Weblog: State-of-the-art NLP fashions from R},
  url = {https://blogs.rstudio.com/tensorflow/posts/2020-07-30-state-of-the-art-nlp-models-from-r/},
  yr = {2020}
}

A easy blood take a look at might spot Parkinson’s years earlier than signs

Science

Dr. Mike

January 29, 2026

A easy blood take a look at might spot Parkinson’s years earlier than signs

Researchers led by a workforce at Chalmers College of Expertise in Sweden have recognized organic markers that seem within the earliest levels of Parkinson’s illness, earlier than main harm happens within the mind. These early modifications depart detectable traces within the blood, however just for a short while. The findings spotlight a vital alternative to each diagnose the illness earlier and discover remedies whereas the mind continues to be largely intact. The researchers imagine blood exams based mostly on this work might start to be examined in healthcare settings inside 5 years.

Parkinson’s illness impacts greater than 10 million individuals worldwide and is taken into account an endemic situation. As populations proceed to age, that quantity is anticipated to greater than double by 2050. Regardless of its rising influence, there’s at the moment no treatment and no extensively used screening methodology that may detect the illness early, earlier than it causes vital and infrequently irreversible mind harm.

New Research Factors Towards Earlier Analysis

The findings had been revealed within the journal npj Parkinson’s Illness by a analysis workforce from Chalmers College of Expertise and Oslo College Hospital in Norway. The research describes main progress towards figuring out Parkinson’s throughout its earliest section, properly earlier than basic movement-related signs seem.

“By the point the motor signs of Parkinson’s illness seem, 50 — 80 per cent of the related mind cells are sometimes already broken or gone. The research is a vital step in direction of facilitating early identification of the illness and counteracting its development earlier than it has gone this far,” says Danish Anwer, a doctoral pupil on the Division of Life Sciences at Chalmers and the research’s first writer.

A Lengthy and Neglected Early Section

Parkinson’s illness develops slowly. In lots of sufferers, the early section can last as long as 20 years earlier than noticeable motor signs totally emerge. Throughout this time, modifications are already occurring inside cells.

The researchers targeted on two organic processes believed to play a task at this early stage. One is DNA harm restore, the system cells use to detect and repair genetic harm. The opposite is the mobile stress response, a protecting response that helps cells survive by shifting power away from routine duties and towards restore and protection.

Machine Studying Reveals a Distinctive Sample

Utilizing machine studying and different superior analytical strategies, the workforce recognized a definite sample of gene exercise associated to DNA restore and stress response. This sample appeared solely in individuals within the early section of Parkinson’s illness. It was not seen in wholesome people or in sufferers who had already developed motor signs.

“Which means that we have now discovered an necessary window of alternative through which the illness may be detected earlier than motor signs brought on by nerve harm within the mind seem. The truth that these patterns solely present at an early stage and are now not activated when the illness has progressed additional additionally makes it fascinating to concentrate on the mechanisms to seek out future remedies,” says Annikka Polster, Assistant Professor on the Division of Life Sciences at Chalmers, who led the research.

Why Blood-Primarily based Testing Issues

Scientists all over the world have been trying to find dependable early indicators of Parkinson’s illness, together with markers discovered by way of mind imaging and spinal fluid evaluation. Nonetheless, none of those approaches has but led to a validated screening take a look at appropriate for widespread use earlier than signs start.

“In our research, we highlighted biomarkers that seemingly replicate a number of the early biology of the illness and confirmed they are often measured in blood. This paves the way in which for broad screening exams by way of blood samples: a cheap, simply accessible methodology,” says Polster.

Blood Exams May Attain Healthcare Inside Years

The following section of the analysis will concentrate on understanding precisely how these early organic mechanisms work and on growing instruments that make them simpler to detect.

The researchers estimate that inside 5 years, blood exams designed to establish Parkinson’s illness at an early stage might start to be examined in healthcare methods. Over the long term, the findings might also assist the event of remedies aimed toward slowing or stopping the illness.

“If we will research the mechanisms as they occur, it might present necessary keys to understanding how they are often stopped and which medicine is perhaps efficient. This may occasionally contain new medicine, but additionally drug repurposing, the place we will use medicine developed for ailments aside from Parkinson’s as a result of the identical gene actions or mechanisms are energetic,” says Polster.

Extra Concerning the Scientific Article

The research Longitudinal evaluation of DNA restore signature trajectory in prodromal versus established Parkinson’s illness has been revealed in npj Parkinson’s Illness. The authors are Danish Anwer, Nicola Pietro Montaldo, Elva Maria Novoa-del-Toro, Diana Domanska, Hilde Loge Nilsen and Annikka Polster. The researchers work at Chalmers College of Expertise, Sweden, and Oslo College Hospital, Norway.

The analysis has been funded by Chalmers Well being Engineering Space of Advance, Sweden, the Michael J Fox Basis, the Analysis Council of Norway, NAISS (Nationwide Educational Infrastructure for Supercomputing in Sweden) and the Swedish Analysis Council.

Extra About Parkinson’s Illness

Parkinson’s illness is a neurological dysfunction that interferes with the mind’s capacity to manage motion. It progresses slowly and most frequently begins after the age of 55 — 60. Parkinson’s is the second commonest neurodegenerative illness worldwide, after Alzheimer’s illness. Greater than 10 million individuals have been recognized globally, and that quantity is projected to greater than double by 2050.

Sources: The Swedish Parkinson’s Affiliation, The BMJ, world projection research, 2024

Parkinson’s Illness Signs and Development

Early signs

REM sleep conduct dysfunction: The individual acts out desires throughout REM sleep, typically with actions or sounds.
Diminished sense of scent
Constipation
Melancholy
Anxiousness

Motor signs later within the illness

Sluggish actions
Rigidity and instability
Tremors
Involuntary muscle contractions

1...313233...324 Page 32 of 324

Producing Extremely Efficient Decks for My Knowledge Science Class

Principled Coarse-Grained Acceptance for Speculative Decoding in Speech

Apiiro’s Guardian Agent guards towards insecure AI code

DHS is utilizing Google and Adobe AI to make movies

Why is Trump on the verge of bombing Iran once more?

What is definitely occurring with Iran’s nukes?

How harmful may an Iranian counterattack be?

Polar bears are getting fatter within the fastest-warming place on Earth

The northern lights, fjords and glaciers: Svalbard and Tromso, Norway

1. What are social media initiatives for college kids?

2. Are these social media challenge concepts appropriate for inexperienced persons?

3. Which expertise can college students be taught from social media initiatives?

4. Can these initiatives be used for varsity or faculty assignments?

5. Do college students want superior instruments to finish these initiatives?

Unified Cisco Catalyst Middle Monitoring with CCGM APIs

Simplify Multi-Cluster Administration with Cisco Catalyst Middle World Supervisor

Construct a Actual-Time Cisco Catalyst Middle Monitoring Dashboard Utilizing CCGM APIs

Why Attend This Cisco Stay DevNet Labs Session?

Session Particulars

Conclusion

State-of-the-art NLP fashions from R

Introduction

Conditions

Template

Information preparation

Information enter for Keras

Conclusion

Corrections

Reuse

Quotation

A easy blood take a look at might spot Parkinson’s years earlier than signs