AI/ML News & Innovations Hub

Fair comparison method for a biased physics-based model and its ML-correction version

I'm working with two prediction models: A calibrated physics-based model that consistently overestimates and has a fixed bias. An XGBoost model that predicts the error of the physics model to create corrected predictions. (The dataset contains the physics model's predictions and the target variable. We then predict the error between the target variable and the physical models prediction) I'm trying to fairly compare these models, but I'm unsure about the best approach. Simply using MAPE for both doesn't seem meaningful due to the known bias in the physics model. Here's what I'm considering: For the physics model: Calculate errors: errors = physics_model_prediction - actual_value Compute mean error: mean_error = errors.mean() Adjust errors: adjusted_errors = errors - mean_error Calculate MAPE: MAPE(adjusted_errors, actual_value) For the corrected model: Calculate MAPE(corrected_prediction, actual_value) Is this approach valid? The residuals' distribution in the correction model is slightly skewed but close to normal, tending to overestimate at certain values. But I'm wondering if this method is statistically sound. If there are better ways to compare the improvements I would appreciate any insights!

Why using mutual information is allowed for feature selection if depends on the "scale" of entropies?

It is common to use mutual information as feature selection method. However, I fail to see why this is the case, since the mutual information $I(X, Y)$ depends on both entropies $H(X)$ and $H(Y)$ via the formula : $$ I(X, Y) = H(X) + H(Y) - H(X,Y)$$ meaning that comparing $I(X_i, Y)$ and $(X_j, Y)$ as a measure for selecting between $X_i$ and $X_j$ is not straightforward since the measure can be bloated by the marginal entropies. It is like selecting between $X_i$ and $X_j$ based on the covariance with $Y$ instead of correlation. The only way I can think that such a comparison is allowed is due to the equivalent formula: $$I(X, Y) = H(Y) - H(Y|X)$$ As the first term $H(Y)$ is the same for all $X_i$ then the ramking shouldn't depend on the "scale" of $H(Y)$ . Is that correct or am I missing something?

Introducing Anyscale’s Unified Log Viewer

Deploy Ray Serve with up to 50% fewer nodes using Anyscale Replica Compaction

Making Insights-Driven Decisions in an Ecosystem of Ecosystems

Our approach to securing our cloud environments

Les sciences du numérique à la conquête du ciel et de l’espace

Les sciences du numérique à la conquête du ciel et de l’espace mtestari mar, 07/09/2024 - 15:23 Après le lancement réussi d’Ariane 6 en juillet 2024 à Kourou, et alors que le nombre de voyageurs aériens atteint de nouveaux records en 2025, les sciences et technologies du numérique revêtent désormais une importance cruciale dans les domaines de l’aérospatial et l’aéronautique. © Pixabay/ Photo V. Stefanov Ces deux secteurs, très présents en Nouvelle-Aquitaine et en Occitanie, couvrent également des enjeux sociétaux et économiques significatifs. C’est tout un écosystème, opéré par Aerospace Valley , premier pôle de compétitivité européen, qui participe à l’étude, la conception, la fabrication et la commercialisation de ces technologies. Dans ce contexte, les équipes du Centre Inria de l’université de Bordeaux offrent à leurs partenaires industriels et académiques, les outils et les connaissances nécessaires pour renforcer la sécurité, la compétitivité et la décarbonation des systèmes, en s’appuyant sur leurs expertises telles que la modélisation, la simulation et la cryptographie . La conception des systèmes aéronautiques : un challenge scientifique pour chaque composante Il est primordial de développer des produits (aéronefs, avions, drones, lanceurs de satellites) les plus performants possible d’un point de vue du service rendu que de l’optimisation de la ressource exploitée. Grâce à la modélisation et à la simulation, Inria contribue à la création de modèles précis et réali…

Using conditional probability as an estimate in a loss function

I have a rather large ML framework that takes multiple conditional probability terms that are computed via classifiers/neural networks. This arbitrary loss function is computed via a function: loss_value = arbitrary_loss(probability1, probability2, ..., P(Y|Z)) I wish to have an end-to-end framework that computes everything and trains everything together. So I do not want to have an independently trained classifier. Say at some point I develop some intermediate values (embeddings) Z from the input samples X. I wish to model the conditional probability P(Y|Z) via an MLP softmax layer. This term P(Y|Z) is then estimated and plugged into the final loss which is the sum and product of other probabilities. P(Y|Z) = MLP(input_Z) #probability given input Z over labels My issue is that if I simply take the value of the softmax layer to estimate this probability and plug it in, at no point are the true labels taken into account for a supervised machine learning problem. How can I fix this without modifying the final loss function? TLDR: I need a probability term P(Y|Z) modeled via an MLP softmax layer to be used in a complex arbitrary loss function. How do i ensure this term is accurate via the true label values, so that it can be used in the final loss?

Understanding Empirical Bayes

I am trying to understand the basics of empirical bayes. I found myself struggling a lot to understand this so I tried to create a toy example involving the estimation of the success probabilities for a coin. 1) Traditional Bayesian Analysis (Beta-Binomial Conjugacy): Consider a sequence of $n$ coin tosses where we observe $X$ heads. The likelihood of observing $X$ heads given the true probability of heads $\theta$ is binomial: $$ P(X|\theta) = \binom{n}{X} \theta^X (1-\theta)^{n-X} $$ Chose a Beta prior distribution for the probability of heads $\theta$ : $$ P(\theta) = \text{Beta}(\alpha, \beta) = \frac{\theta^{\alpha-1} (1-\theta)^{\beta-1}}{B(\alpha, \beta)} $$ Using Bayes' theorem, the posterior distribution is proportional to the likelihood times the prior: $$ P(\theta|X) \propto P(X|\theta) P(\theta) $$ Substituting the binomial likelihood and the Beta prior, we get: $$ P(\theta|X) \propto \theta^X (1-\theta)^{n-X} \cdot \theta^{\alpha-1} (1-\theta)^{\beta-1} $$ $$ P(\theta|X) \propto \theta^{X+\alpha-1} (1-\theta)^{n-X+\beta-1} $$ Thus, the posterior distribution is also a Beta distribution: $$ P(\theta|X) = \text{Beta}(X + \alpha, n - X + \beta) $$ 2) Empirical Bayes: It seems like in Empirical Bayes, the parameters of the priors are estimated from the data instead of being chosen before hand. To me it makes more sense to use Method of Moments (instead of MLE) to estimate the parameters $\alpha$ and $\beta$ of the Beta prior. The sample mean $\hat{\theta}$ and sampl…

What's the relation between Generalized Policy Iteration (GPI), Actor-Critic, and Q-learning methods?

It seems to me that Generalized Policy Iteration (GPI) and Actor-Critic are the same, and Q-learning methods are a separate family of algorithms. I think both GPI and Actor-Critic describe the iterative process of policy evaluation (critic) and policy improvement (actor), while Q-learning is only bootstrapping using the Bellman optimality equation. To elaborate on my understanding: Policy evaluation (critic) is done via Monte-Carlo or temporal difference methods, including function approximation if necessary, and policy improvement (actor) can be done by being greedy (tabular case) or using the policy gradient theorem (large state space). Q-learning is not doing any of these. It's just trying to estimate the $Q^\ast$ using the Bellman optimality equation by iteratively fitting the Bellman optimality equation for Q value. I would appreciate it if anyone can confirm whether my understanding is correct or give a more systematic/precise summary of the taxonomy of online RL algorithms.

Extrinsic Hallucinations in LLMs

Hallucination in large language models usually refers to the model generating unfaithful, fabricated, inconsistent, or nonsensical content. As a term, hallucination has been somewhat generalized to cases when the model makes mistakes. Here, I would like to narrow down the problem of hallucination to cases where the model output is fabricated and not grounded by either the provided context or world knowledge. There are two types of hallucination: In-context hallucination: The model output should be consistent with the source content in context. Extrinsic hallucination: The model output should be grounded by the pre-training dataset. However, given the size of the pre-training dataset, it is too expensive to retrieve and identify conflicts per generation. If we consider the pre-training data corpus as a proxy for world knowledge, we essentially try to ensure the model output is factual and verifiable by external world knowledge. Equally importantly, when the model does not know about a fact, it should say so. This post focuses on extrinsic hallucination. To avoid hallucination, LLMs need to be (1) factual and (2) acknowledge not knowing the answer when applicable.

How do you save a stable diffusion model locally for later us?

I am new to ML and plan to use KerasCV stabledifussion model to generate images from text. The example on the KerasCV website is straightforward but I could not find a way to save the model locally for later use. I also noticed that the library connects to hugging face to download encoder and diffusion model. Could you please point me to the right direction to do this locally? I would like all the model and its parameters to be local and I will be using it in a server. Also, if you have experience running such a model/server on the could, I would appreciate your guidance on the best approach wrt costs. Should I upload everything and store the whole data on the cloud or load it from hugging face? Which one would make more sense for cloud applications?

New paper: AI agents that matter

Rethinking AI agent benchmarking and evaluation

An Introduction to the Code of Practice for General-Purpose AI

Last updated: 14 August 2025. As AI Act implementation gradually unfolds, it is important to understand the different mechanisms of enforcement included in the Regulation. One of the most important is the general-purpose AI Code of Practice, which was developed by the AI Office and a wide range of stakeholders. This summary, detailing the Code […]

Predicting Values with Bayesian Neural Network

I want to use a Bayesian Neural Network for a regression task. To do that I converted a BNN from this paper to Python 3. The provided training script runs and I receive a pickle file, which I want to use to predict a value for my regression. Event though the loss for the training doesn't really go down, but this seems also be the case for the models used in the paper. The input of my network is a vector with 59 features, where one feature is my target variable. It's pretty similar to the boston housing dataset. The MLP inside the BNN returns two vales: $\mu$ and $\sigma$ . If I understand BNNs correctly $\mu$ and $\sigma$ refer to the mean and standard deviation of my input and I should use $\mu$ to predict my values. But when I use input vector I always receive a value that is roughly around the mean of my target variable. So lets say the mean is 62. When I insert a vector from my test set, that should return the target variable 142, I will get 61. When I do the same with a vector, that should return 21, I will get 63. So my predicted values are always around the mean of the target variable. What do I have to do, to get real predictions from my input data? I never worked with BNNs before and it's quite hard to find any resources on how I should model my network. Maybe someone has a nice tutorial to learn how I can use this model to predict my target variable. I have to use this class and can't use Pyro or any other framework currently, because otherwise I can't use the othe…

Forecasting time series using simulations

Suppose we have a stationary time series $x_{1}, x_{2}, ..., x_{T}$ . Goal is to forecast up to $T+h$ , i.e., forecast $x_{T+1}, x_{T+2}, ..., x_{T+h}$ . Forecasting methodology: Using econometric techniques one can try to fit a model, which describes the data generating process of $\{x\}_{t}$ (e.g., ARIMA or other model). Then, having estimated parameters of the model one can simulate using forward Monte Carlo technique many paths up to $T+h$ . Using simulated paths, one can construct an empirical PDF for each point in time in the future, i.e., $f_{T+1}(x), f_{T+2}(x), ..., f_{T+h}(x)$ . As a forecast for $x_{T+1}, x_{T+2}, ..., x_{T+h}$ , take mean of $f_{T+1}(x), f_{T+2}(x), ..., f_{T+h}(x)$ , respectively. Question: Is this kind of method acceptable or widely used for stationary process forecasting? What are the main problems and assumptions of this method? Also, I assume that forecasts will be the same as the current value, i.e., $E(x_{T+i}|x_T)=x_{t}$ for any $i$ . Is this assumption correct?

Recap: Square Unboxed 2024

Top highlight's from this year's event

Announcing the Winners of the Square Local Community Hackathon

Building apps to connect businesses with their local community

How to actually use the Empirical Influence Function for BCa Bootstrap Intervals?

In the course of seeking reassurance for another part of a hobby analysis* I found a stack answer which mentioned The Jackknife, the Bootstrap and Other Resampling Plans (Efron, 1980). Having managed to find the paper online, the bit about non-parametric bias and skew adjustments to the bounds of bootstrap CIs caught my eye. The problem is that I don't really understand how to do it as described in that paper . Efron (1980) as I am reading it suggests that the core ingredients of the BCa CI method are the $U_i$ and $a$ values. The derivation of the latter is simple given you know the $U_i$ with formula (pg. 21): $$ a \doteq \frac{1}{6}\frac{\sum\limits_{i=0}^n U_i^3}{\left( \sum\limits_{i=0}^n U_i^2 \right)^{\frac{3}{2}}} $$ The $U_i$ have a more complicated formula, coming from what's termed the empirical influence function (also pg. 21): It is this formula for the $U_i$ that I don't know how to work with. Clicking around in some more stack questions, I found a reference to Bootstrap Methods and their Application (Davison and Hinkley, 1997) but I couldn't find that online. Nevertheless, I did find a slideshow by Davison on ResearchGate based on it which proposes that the $a$ value can be jackknifed like so (I've changed the subscript $j$ to $i$ to match Efron, 1980): $$ l_i \approx l_{jack,i} = (n-1)(\hat{\theta} - \hat{\theta}_i)$$ where $n$ is the number of observations in the original sample. I have tested my understanding of the jackknife estimate of $a$ against the $U_…

End-to-end LLM Workflows Guide

How are perplexities over multiple instance aggregated?

The perplexity of the $i^{th}$ token in the $k^{th}$ sequence is $$ P_{ki} = \frac{1}{p(t_{ki})} $$ The perplexity aggregated for the $k^{th}$ sequence is then $$ P_{k} = \left(\prod_{i=1}^N P_{ki}\right)^{1/N} \\ = \left(\prod_{i=1}^N \frac{1}{p(t_{ki})} \right)^{1/N} $$ which is the geometric mean of the perplexities of the tokens. This makes sense as we are essentially taking the multiplicative inverse of the probability that the model got the whole sequence correct. Now my question is how to aggregate the perplexities of several sequences. It seems from various places, including the Hugging Face Tutorial , I see that the prescription is to take the arithmetic mean of the perplexities of sequences $$ P = \frac{1}{m} \sum_{k=1}^m P_k $$ I am not quite understanding what it means to take the average of 1/probabilities. What is this actually capturing?

Experiments in Weak-to-Strong Generalization

Writing up results from a recent project

Moderator Analysis in Cox Regression

Imagine an RCT with two groups with a time-to-event endpoint. The (pre-specified) strategy for analysing this trial is a Cox Regression using common covariate adjustment to reduce outcome heterogeneity. So far, so good. Now I want to know whether my treatment effect is moderated by level of education (3 levels). If I used the pre-specified cox regression with covariate adjustment and integrated the treatment*education interaction, the SEs are getting very large. What may the reasons be? Is my sample size (120 per group) too little and should I reduce the covariates for which I adjust? Edit: I controlled for 4 covariates. There were 76 events in the IG and 91 events in the CG. More specifically: Education level1: IG: 13 CG: 18. Education level2: IG: 32 CG: 28. Education level3: IG: 31 CG: 45

Free Form Least-Squares Concept Erasure Without Oracle Concept Labels

Achieving even more surgical edits than LEACE without concept labels at inference time.

DDPG model outputting a fixed action at every timestep

I am trying to create a Car Following model, for which i am using DDPG. My action is acceleration bounded in a range of [-3,3] m/s2. While training the model, for every state it gives a single acceleration value i.e. 3 (or sometimes -3). It can be clearly seen that my actor is performing really bad. What can be done to resolve this issue?

Why work at the EU AI Office?

It's probably not for everyone, but there are a lot of great reasons to consider, including the potential to have an impact on AI governance worldwide, leveraging the first-mover advantage, and more.

Would the DDPG algorithm still function effectively if some transitions stored in its replay buffer are generated by a completely unrelated policy?

Let's hypothesize a scenario where some of the records ( s i , a i , r i , s i+1 ) in the replay buffer are generated by another completely unrelated random policy. If the DDPG algorithm still samples random minibatches from this buffer for learning as usual, would the learning process proceed successfully? Actually, there's a pre-processing stage before training DDPG in my online learning application, where another module learns the safe action range. I wonder if the transition records obtained during this stage can be used to pre-train the DDPG agent.

Q&A: Discover what’s new in the Mobile Payments SDK beta

Learn about how KIOS, a kiosk software, conducted early stage testing of the SDK and their integration experience.

Cross-modal Search for E-commerce: Building and Scaling a Cross-Modal Image Retrieval App

Minimum Survey Respondents for Hypotheses I want to answer

I am an eager statistics noob, so apologies on the basic questions. I am looking to send out a survey to a population that is large (people with diabetes who take insulin in the US) (~ 8.4 million) Based on this I am aiming for 96+ survey responses overall (with a margin of error of +- 10%), with a goal of getting closer to 300 responses. I have some hypotheses I want to look at in the survey results and will be doing some statistical testing One hypothesis is "Hypothesis: There is a significant difference in rating of importance of (xyz user need, just giving a fake example - being able to quickly calculate amount of insulin for a meal) based on time the user has been using insulin" This would be a demographic question I collect (How long have you been using insulin for?), with a range of responses in months. Since importance is a rating on a likert scale (ordinal), the groups are different, and I want to understand difference, I was thinking of doing the Mann-Whitney test. I specifically would like to understand if longer time using insulin would impact rating of importance on this one user need. My question: Would I need a minimum of 96 respondents who answer in each bucket (for example 1 year or longer and 1 year or shorter) to aim for a stat sig response or would I just need to be looking at 96 overall for all respondents who answer this q. Also, if you have feedback on my approach feel free to weigh in.

Is reinforcement learning suitable for application automation?

I have basically automatised the use of an app through the use of OCR and computer vision. So basically when a word or an image is detected it will perform a certain action. When that action is successfully completed it will go to the next state. Now I want to try basically with a more "heuristic" approach and I thought about reinforcement learning. Why? Because I am aiming to build a tool that basically understand automatically what actions to perform in a certain state. But I have a doubt. Even though I don't need to declare an association like this (it would beat the purpose of deep reinforcement learning or deep learning in general): if(state.MENU_VIEW) clickManager.clickOnFolder(); ... I still need to define the states, the actions and the reward. Meaning I would need to instruct my app that when the OCR result is "Open Folder" it means the state I am in is MENU_VIEW. I simply wouldn't tell my app what action to perform in a that state. Am I correct? What I am trying to say is: how exactly could I make it so that the states (and maybe also the actions?) are generated automatically? The reward in this case scenario would be basically the folder being opened successfully.

Probability interpretation of attention mechanism in Seq2Seq

I have ready many explanations of the seq2seq model. In my opinion, however, it is really like a robot that might say something correctly, but doesn't really understand it, just as is true with an LLM generally. In my opinion, the correct way to describe Seq2Seq and similar NLP models should start from a probability view. My probability view is very simple; the output of the encoder is a representation of the probability distribution of the next word. In each step of the Decoder, it just modifies the distribution based on each word it predicted from the distribution and outputs the modified distribution. It then does this repeatedly. Assuming this probability view is correct, how could we explain the attention mechanism used in Seq2seq?

Providing lawyers with a leg-up in navigating AI for legal practice

AISG and the Singapore Academy of Law (SAL) will work together to support the legal profession as it faces a time of transformation, precipitated by the rise of Generative...

IBM and AI Singapore Ink MOU to Collaborate on First Large Language Model developed with Southeast Asian context

Ahead of ATx Enterprise 2024, IBM announced the signing of a memorandum of understanding (MOU) with AI Singapore (AISG) on 28 May 2024 to collaborate on the testing of...

VINC-S: Closed-form Optionally-supervised Knowledge Elicitation with Paraphrase Invariance

Writing up results from a project from Spring 2023

We Pre-Trained Stable Diffusion Models on 2 billion Images and Didn't Break the Bank - Definitive Guides with Ray Series

Using F-test for restricted model vs unrestricted, wrong answer

studying Econometrics I come across this question for which I cannot find the right answer: Assuming a percentage of household's expenses on food is linearly dependent on the total expenditure and size of household: $$\text{wfood}_i = \beta_1 + \beta_2 \text{totexp}_i + \beta_3 \text{size}_i$$ Using F-test for restricted and unrestricted models test a hypothesis $H_0$ that coefficients $\beta_1, \beta_2, \beta_3$ do not depend on the $\text{sex}_i$ . The alternative hypothesis is that at least one of $\beta_1, \beta_2, \beta_3$ vary. I use $$F=\frac{(RSS_R-RSS_{UR})/r}{RSS_{UR}/(n-k_{UR})} \sim F_{r, k_{UR}}$$ to test restricted model (first one) against unrestricted (adding $+\beta_4\text{sex(woman)}_i$ ), but get the wrong value of F-statistic (I have to choose one of the variants, and there's no such value there; expected value is $2.4$ or $24.4$ ). How do I do that: library(Ecdat) data("BudgetFood") b I appreciate any ideas. Thanks.

Structure and parameters of modern AI-based chess engines

I understand some recent chess engines (like AlphaZero or MuZero) are based on neural networks. This question is not specific to chess, any other game (e.g. go) would do, but I keep chess for concreteness. I am not interested in how these engines are trained but in what they "look like" after training. (For concreteness, it suffices for me that they give a "score" for a given board position.) Somewhat similar questions have been asked here but I can't find what I need. Is it correct to view them as a neural network? Or is there some other major component like search trees or other type of algorithm. I'm assuming that answer here is "yes" in an appropriate sense. Note: The AlphaZero paper talks about various forms of tree search. It is not clear to me how they are used to play, and from the appendix apparently some other engines use very limited search, like depth 2? That would be already interesting to me. What is the depth of the network? Any reference to precise parameters? What is the input and output format of these networks? Is the input just an array with pieces positions (excluding three-fold repetition rules and similar secondary issues). What is the output? Some info is here but I can't quite understand how it works specifically e.g. for chess. What activation functions are used?

Processing 2 Billion Images for Stable Diffusion Model Training - Definitive Guides with Ray Series

Prove Wald statistic = number of linear restrictions times F statistics

I'm considering $F[J,n-k]= \displaystyle \frac{(e_{*}^{'}e_{*}-e'e) \backslash J}{e'e\backslash (n-k)}$ , where $J$ stands for number of restrictions. I want to prove $W=(Rb-q)'(Rs^2(X'X)^{-1}R')^{-1}(Rb-q)=JF$ , where $s^2=\displaystyle{\frac{e'e}{n-k}}$ . So it is enough to show $(e_{*}^{'}e_{*}-e'e)$ is $(Rb-q)'(R(X'X)^{-1}R')^{-1}(Rb-q)$ , $R$ stands for the restrictions. In Greene's textbook, he states this, but I don't know how to derive it. In a particular, probably simpler example, consider a partitioned linear regression model $Y=X_1\beta_1+X_2\beta_2+\epsilon$ , and $H_0$ : $\beta_2=0$ , so $Rb-q=b_2$ which is an OLS estimator of $\beta_2$ . $\beta_1$ is $k_1\times 1$ and $\beta_2$ is $k_2\times 1$ so $J$ should be $k_2$ .We want $k_2F=W$ . But I don't know how to derive it even in this example. Could you please give a proof in example and then maybe in general case?

What properties must be verified for a simple OLS regression model with time series?

Let's say I have 3 time series variables, $(X_t)$ , $(Y_{t})$ , $(Z_{t})$ and I estimate the following model with OLS estimator (I believe this form is called ARDL) : $$X_{t} \quad = \quad \alpha_1 X_{t-1} \quad + \quad \alpha_2 Y_{t} \quad + \quad \alpha_3 Z_{t-1} \quad + \alpha_4Z_{t-2} \quad + \quad \varepsilon_t$$ where $\varepsilon_t$ is the error term. What properties should be verified for this regression to hold ? Here is what I thought should be checked : $(X_t)$ , $(Y_{t})$ and $(Z_{t})$ are stationary time series no multicolinearity between $(X_t)$ , $(Y_{t})$ and $(Z_{t})$ i.e. correlations between variables are sufficiently weak. the error $\varepsilon_t$ is homoskedastic : $var(\varepsilon_t) = \sigma$ where $\sigma$ does not depend on $t$ the error $\varepsilon_t$ has no autocorrelation : $cov(\varepsilon_t,\varepsilon_{t-k})=0 \quad \forall (t,k) \in \Bbb{N}^2 $ the error $\varepsilon_t$ is centered : $\mathbb{E}(\varepsilon_t) = 0 $ the error $\varepsilon_t$ is stationary (but is that not already implied by what's above ?) the error $\varepsilon_t$ is normaly distributed (optional) the AR(1) and AR(2) processes of $(X_t)$ and $(Z_t)$ have coefficients smaller than 1 in absolute value in order to be stationary : $ |\alpha_1| \in \left]0;1\right[ $ $ |\alpha_3| \in \left]0;1\right[ $ $ |\alpha_4| \in \left]0;1\right[ $ Is there something I missed ? Thanks a lot !

How to get Complexity per Layer, Sequential Operations and Maximum Path Length in CNN architecture?

In the paper Attention is all you need , here is Table 1, can someone explain what architecture is referred to in the "Convolution" row and hence describe the other 3 columns in it? The other ones are pretty clear, for example Recurrent, takes $O(d^2)$ operations in one time-step to multiply the hidden state with the weight matrix, and there are $n$ such time-steps, which need to be done sequentially ( $O(n)$ ), also making the information from the first and last tokens in the sentence to travel $O(n)$ steps. For Self-attention, every word in the sentence, attends to every other, so $O(n^2)$ pairs, dot product attention taking $O(d)$ operations. Clearly, the path length is $O(1)$ and there are no sequential operation, all these $O(n^2)$ pairs can be computed independently. The "Convolution" row is not clear to me!

Why some papers focus on constructing large dataset from real robots, instead of simulations?

Recently, I have seen papers about large datasets for robotics such as DROID( https://droid-dataset.github.io/ ) or Open X-Embodiment( https://robotics-transformer-x.github.io/ ). As I see, the datasets are specific to some types of robots(although X-Embodiment allows one robot to learn from another robot's data) and environments. If one wants to add another robot into the dataset, they have to do all data sampling again, which is quite expensive. Some environments might be difficult to reproduce, especially as they collected data from all the labs in the world. I am wondering: why don't they instead set up data collection procedure on simulation? it will make the data collection way cheaper. When they want to add a new robot and collect data with the same tasks and environments like other robots, they can do it easily. It is also easy to add a new task and collect data from all robots/environments. Then, why they collect data in real world while giving up on such reproducibility/extensibility? Is Sim2Real that bad, even if it can collect way more samples easily?

Financial Market Applications of LLMs

The AI revolution drove frenzied investment in both private and public companies and captured the public’s imagination in 2023. Transformational consumer products like ChatGPT are powered by Large Language Models (LLMs) that excel at modeling sequences of tokens that represent words or parts of words [2]. Amazingly, structural

Measuring personal growth

My founder friends constantly think about growth. They think about how to measure their business growth and how to get to the next order of magnitude scale. If they’re making $1M ARR today, they think about how to get to $10M ARR. If they have 1,000 users today, they think about how to get to 10,000 users. This made me wonder if/how people are measuring personal growth. I don’t want to use metrics like net worth or the number of followers, because that’s not what I live for. After talking with a lot of friends, I found three interesting metrics: rate of change, time to solve problems, and number of future options. Some friends told me they find this blog post mildly sociopathic. Why do I have to measure everything? Life is to be lived, not to be measured. As someone lowkey fascinated by numbers, I don’t see why measuring and living have to be mutually exclusive – measuring often helps me live better – but I see where they come from. This post is more of a thought exercise than a rigorous experiment. Rate of change I have this theory that life has a circadian rhythm. Every 3-6 years, you become a different person. You work on different problems. Your lifestyle changes. The people you hang out with are different. If you haven’t caught up with a friend in 5 years, you might no longer have anything in common. It’s not a coincidence that schools are structured into chunks of 3-6 years. Looking back, I realized that every 3-6 years, my life completely changed. From grade 3 to grad…

Pile-T5

Trained T5 on the Pile

Developing Advanced RAG Systems with Qdrant Hybrid Cloud and LangChain

LangChain and Qdrant are collaborating on the launch of Qdrant Hybrid Cloud , which is designed to empower engineers and scientists globally to easily and securely develop and scale their GenAI applications. Harnessing LangChain’s robust framework, users can unlock the full potential of vector search, enabling the creation of stable and effective AI products. Qdrant Hybrid Cloud extends the same powerful functionality of Qdrant onto a Kubernetes-based architecture, enhancing LangChain’s capability to cater to users across any environment.

Diffusion Models for Video Generation

Diffusion models have demonstrated strong results on image synthesis in past years. Now the research community has started working on a harder task—using it for video generation. The task itself is a superset of the image case, since an image is a video of 1 frame, and it is much more challenging because: It has extra requirements on temporal consistency across frames in time, which naturally demands more world knowledge to be encoded into the model. In comparison to text or images, it is more difficult to collect large amounts of high-quality, high-dimensional video data, let along text-video pairs. 🥑 Required Pre-read: Please make sure you have read the previous blog on “What are Diffusion Models?” for image generation before continue here.

Red Hat OpenShift and Qdrant Hybrid Cloud Offer Seamless and Scalable AI

We’re excited about our collaboration with Red Hat to bring the Qdrant vector database to Red Hat OpenShift customers! With the release of Qdrant Hybrid Cloud , developers can now deploy and run the Qdrant vector database directly in their Red Hat OpenShift environment. This collaboration enables developers to scale more seamlessly, operate more consistently across hybrid cloud environments, and maintain complete control over their vector data. This is a big step forward in simplifying AI infrastructure and empowering data-driven projects, like retrieval augmented generation (RAG) use cases, advanced search scenarios, or recommendations systems.

Qdrant Hybrid Cloud and DigitalOcean for Scalable and Secure AI Solutions

Developers are constantly seeking new ways to enhance their AI applications with new customer experiences. At the core of this are vector databases, as they enable the efficient handling of complex, unstructured data, making it possible to power applications with semantic search, personalized recommendation systems, and intelligent Q&A platforms. However, when deploying such new AI applications, especially those handling sensitive or personal user data, privacy becomes important. DigitalOcean and Qdrant are actively addressing this with an integration that lets developers deploy a managed vector database in their existing DigitalOcean environments. With the recent launch of Qdrant Hybrid Cloud , developers can seamlessly deploy Qdrant on DigitalOcean Kubernetes (DOKS) clusters, making it easier for developers to handle vector databases without getting bogged down in the complexity of managing the underlying infrastructure.

Enhance AI Data Sovereignty with Aleph Alpha and Qdrant Hybrid Cloud

Aleph Alpha and Qdrant are on a joint mission to empower the world’s best companies in their AI journey. The launch of Qdrant Hybrid Cloud furthers this effort by ensuring complete data sovereignty and hosting security. This latest collaboration is all about giving enterprise customers complete transparency and sovereignty to make use of AI in their own environment. By using a hybrid cloud vector database, those looking to leverage vector search for the AI applications can now ensure their proprietary and customer data is completely secure.