14

The article discusses the Adam optimizer, a popular algorithm in deep learning known for its efficiency in adjusting learning rates for different parameters.

Unlike other optimizers like SGD or Adagrad, Adam dynamically changes its step size based on the complexity of the problem, analogous to adjusting stride in varying terrains. This ability to adapt makes it effective in quickly finding the minimum loss in machine learning tasks, a key reason for its popularity in winning Kaggle competitions and among those seeking a deeper understanding of optimizer mechanics.

16
submitted 6 months ago by Chruesimuesi@feddit.ch to c/buenzli@feddit.ch
[-] Chruesimuesi@feddit.ch 24 points 6 months ago

I'm sorry to hear about your difficult experience and the impact it's had on you. It's obvious that this situation left a deep scar on your soul, and I applaud you for seeking therapy and getting sober.

I, personally don't know you nor her, nor was I present when all this happened but I think it's important to remind yourself that there are always two sides to a story. While your intentions may have been innocent, her perception of the interactions might have been different. It's possible that what felt like friendly conversation to you was perceived as uncomfortable or intrusive to her. This doesn't necessarily make anyone the "bad guy" – it's just a reminder of how complex human interactions can be and how two people can interpret the same situation very differently.

Regarding your self-perception and fear about future relationships, it's crucial to understand that one incident doesn't define who you are or dictate your future. People grow and change, especially when they actively work on themselves as you have. Being sober and in therapy are important steps towards understanding yourself and learning how to build healthy relationships.

Regarding how she chose to handle the situation, it's important to acknowledge that her actions, whether perceived as right or wrong, are beyond your control. While it's possible that her intentions were not entirely good-hearted, focusing on this aspect might not be constructive for your own healing journey. You cannot change her actions or her perception of the events, only how you respond and learn from the experience. This is part of accepting the past and focusing on your own growth and future.

I want to stress, that self-forgiveness is a vital part of healing. Continually hating yourself for past mistakes is not productive. Recognizing your growth and the efforts you've made to improve is important. You're not the person you were six years ago.

Finally let me tell you that everyone deserves love and the chance to enter into a healthy relationship. This experience doesn't change that. I highly recommend discussing your feelings with your therapist, who can provide more personalized guidance and support.

And as last two cents: remember, growth often comes from challenging experiences. You're on the right path by acknowledging the past, learning from it, and making positive changes. Keep moving forward!

I hope you find something useful in my babbling and wish you a wonderful day 🙂

[-] Chruesimuesi@feddit.ch 26 points 10 months ago

Big beards can alter facial recognition and obscure expressions, making someone look more unpredictable or wild to observers.

38
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
258
submitted 10 months ago by Chruesimuesi@feddit.ch to c/aww@lemmy.world
79
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
116
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
132
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
118
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
35
submitted 10 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
18

Large language models (LLMs) are data-efficient but their size makes them difficult to deploy in real-world scenarios.

"Distilling Step-by-Step" is a new method introduced by Google researchers that enables smaller models to outperform LLMs using less training data. This method extracts natural language rationales from LLMs, which provide intermediate reasoning steps, and uses these rationales to train smaller models more efficiently.

In experiments, the distilling step-by-step method consistently outperformed LLMs and standard training approaches, offering both reduced model size and reduced training data requirements.

126
submitted 11 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
31
submitted 11 months ago by Chruesimuesi@feddit.ch to c/gifs@lemmy.world
[-] Chruesimuesi@feddit.ch 11 points 11 months ago

But depending on your current knowledge and future accidents it might be more useful 😉

[-] Chruesimuesi@feddit.ch 30 points 11 months ago

I also don't have anything to add other than that I really appreciate comments who pay respects to details of other comments. I don't know, just makes me happy, so thank you for that!

Oh, and fuck this fucking asshole named Putin, may he die a painful and slow death.

[-] Chruesimuesi@feddit.ch 14 points 1 year ago

This comment was so wholesome it made my day 🥰

[-] Chruesimuesi@feddit.ch 17 points 1 year ago

I guess it could cause a severe seizure or even catastrophic neural failure. The synchronous firing would disrupt the balance of excitatory and inhibitory signals in the brain, potentially damaging neurons and leading to a loss of normal brain function.

The result might include a loss of consciousness, significant cognitive impairments, or even death, depending on the extent and duration of the event.

But then again, I have no idea, since I ain't an expert in that field and in fact am actually just a dog who figured out how to use a computer 🙃

[-] Chruesimuesi@feddit.ch 23 points 1 year ago

It's called Canvas

There is also a community for it :)

[-] Chruesimuesi@feddit.ch 71 points 1 year ago

I don't think so. If you look at Mastodon it could actually keep most of its users and still seems to be growing.

Of course I don't know what the future holds for us.

[-] Chruesimuesi@feddit.ch 43 points 1 year ago

An academic book about emojis that can't include emojis? That's ironic and frustrating. Makes me sad that we live in a world where copyright hinders education and discussion 🙈

Here is a Tl;Dr for the ones who don't want to click the link:

Oxford professor Jieun Kiaer published an academic book called "Emoji Speak: Communications and Behaviours on Social Media," exploring how emojis are used across different cultures and ages, and considering their future in digital communication.

Although the book discusses emojis in detail, Kiaer was unable to include actual images of many emojis due to copyright concerns, despite the fact that these symbols are ubiquitous in social media spaces, which are almost entirely copyright-free.

Instead of using actual emojis, Kiaer hired an artist, Loli Kim, to draw similar representations, illustrating the barriers that exist between the online and offline worlds concerning copyright.

The inability to use emojis in the book, even in an academic context, highlights the complications and absurdity of modern copyright laws, which some argue could have constituted a fair use situation.

[-] Chruesimuesi@feddit.ch 21 points 1 year ago* (last edited 1 year ago)

Some in the AI industry have proposed concepts similar to Moore's Law to describe the rapid growth of AI capabilities.

Although there is no universally accepted law or principle akin to Moore's Law for AI, people often refer to trends that describe the doubling of model sizes or capabilities over a specific time frame.

For instance, OpenAI has previously described a trend where the amount of computing power used to train the largest AI models has been doubling roughly every 3.5 months since 2012.

Source

[-] Chruesimuesi@feddit.ch 58 points 1 year ago

Wait, wtf... Volkswagen killed monkeys in emission tests?

https://www.theguardian.com/business/2018/jan/29/vw-condemned-for-testing-diesel-fumes-on-humans-and-monkeys

Holy fuck you are right. Wtf is wrong with people...

[-] Chruesimuesi@feddit.ch 44 points 1 year ago

Please tell me it's publishing the first version 😘

view more: next ›

Chruesimuesi

joined 1 year ago