Creează un cont pentru a te abona la acest blog și a-i urmări articolele pe pagina Lectură.
-
tags Generative modelling papers (Sohl-Dickstein et al. 2015), (Ho et al. 2020) Principle of diffusion Forward diffusion An image of size \(N\) by \(N\) \(x_0\), which is a vector in \(\mathbb{R}^{N \times N \times c}\) is diffused at each timestep \(t\)…
-
tags Optimization, Algorithm resources Fabian Pedregosa’s series on FW Definition It was originally published in (Frank, Wolfe 1956) and (Jaggi 2013) gives a more recent overview.
-
source (Ainsworth et al. 2022) tags Neural networks TODO Summary This paper introduces various methods for matching and interpolating the weights of multiple neural networks of the same architecture trained from different starting points or data.
-
tags Physics Definition By a complex system I mean one made up of a large number of parts that interact in a nonsimple way. — Herbert Simon, 1962 Bottomless wonders spring from simple rules, which are repeated without end. — Mandelbrot, ~1980 When we talk…
-
tags Economics, Economic liberalism Definition In (Hay 2004), the author gives the following definition of neoliberalism; Economic neoliberalism, I suggest, can be defined in terms of the following traits: A confidence in the market as an efficient mechan…
-
tags Neural networks, Optimization authors Francis Bach, Lénaïc Chizat source Francis Bach’s blog In the rest, we use the mathematical definition of a neural network from Neural networks.
-
tags Optimization Linear programs are problems that can be expressed as \begin{align} & \text{Find a vector} && \mathbf{x} \\ & \text{that maximizes} && \mathbf{c}^T \mathbf{x}\\ & \text{subject to} && A \mathbf{x} \leq \mathbf{b} \\ & \text{and} && \math…
-
tags Neural networks Hopfield networks are a kind of recurrent neural network with binary threshold nodes.
-
tags Life The problem of defining life Definitions of life have always been elusive.
-
tags Emergence, Chaos, Artificial Intelligence resources Wikipedia, (Von Neumann, Burks 1966; Wolfram 2002) Definition A cellular automaton is a computational model defined with respect to a regular grid of individual elements (called cells).
-
tags Economics
-
tags Machine learning Continual learning is a type of supervised learning where there is no “testing phase” associated to a decision process.
-
tags Finance, Economics
-
tags Language resources https://s.ai/nlws/ A fascinating writing system based on glyphs connected to each other to create meaning.
-
tags Applied maths Ramified optimal transport Introduction to ramified optimal transportation
-
tags Algorithm Graham scan is an algorithm to find the convex hull of a set of points in 2D.
-
source (Flageat, Cully 2020) tags ALife 2020, MAP-Elites, Quality diversity Summary MAP-Elites can be problematic in face of uncertainty because: individuals can be unexpectedly lucky the behavior space can be hard to estimate and result in misplacing ind…
-
tags Applied maths Definition The KL divergence is not symmetric.
-
tags ALife 2020, Cellular automata, Autopoiesis source (Cika et al. 2020) Summary This paper is about the possible resistance of GoL patterns to perturbations and the structures that could enable this to happen.
-
Notes on: Safe Reinforcement Learning through Meta-learned Instincts by Grbic, D., & Risi, S. (2020)
source (Grbic, Risi 2020) tags Meta-learning, Reinforcement learning, ALife 2020 Summary In RL an important goal is to find agents that can quickly adapt to changing environments while avoiding unsafe states.
-
tags Emergence, Biological life, ALife 2020 How do organisms store information and are able to pass it down through very profound structural changes (from a caterpillar to a butterfly, when cutting a flatworm in multiple pieces, etc.)?
-
tags Coding Hilbert curves can be used for an interesting trick involving 2D arrays indexing.
-
tags Neural networks source (Greydanus 2020) DONE Summary This paper introduces a minimalist 1D version of the MNIST dataset for studying some basic properties of neural networks.
-
tags NLP
-
tags Statistics A set of numbers satisfies Benford’s law if the leading digits of these numbers occur with a probability logarithmically decreasing with the digit.
-
tags Machine learning paper (Van der Maaten, Hinton 2008) Example: Embedding the vertices of a high dimensional cube in 2D We first create a cube with dimension 12: import imageio import numpy as np import matplotlib.pyplot as plt from sklearn.manifold im…
-
tags Machine learning, Neural networks Adversarial examples in Reinforcement learning Adversarial examples in Computer vision Adversarial examples in NLP A Python library for creating and using text attacks: TextAttack.
-
tags Economics, Europe resources Wikipedia
-
tags Transformers, NLP paper (Radford et al. 2018) Succesors The GPT architecture was improved upon and extended into GPT-2 and GPT-3.
-
tags Transformers, Reinforcement learning paper (Reed et al. 2022) Architecture A standard decoder-only transformer is preceded by an embedding layer that embeds text and images with positional encoding and spatial information if available.
-
tags Transformers, Transformer-XL, NLP paper (Yang et al. 2020) Architecture The model adapts Transformer-XL to be a permutation based language model.
-
tags Transformers, RoBERTa, NLP paper (Conneau et al. 2020) Architecture The model is an extension of RoBERTa that introduces small parameter tuning insights in the context of multilingual applications.
-
tags Transformers, NLP website Wikipedia page for Wu Dao Architecture It is similar to GPT, being a decoder architecture but it applies a different pre-training task.
-
Links Project slides Project report Description The goal of this project was to write a review of adversarial examples in RL as of January 2019.
-
Links Project slides PDF project report
-
Links Project slides PDF project report Github