-
Versatile and Generalizable Manipulation via Goal-Conditioned Reinforcement Learning with Grounded Object Detection
Authors:
Huiyi Wang,
Fahim Shahriar,
Alireza Azimi,
Gautham Vasan,
Rupam Mahmood,
Colin Bellinger
Abstract:
General-purpose robotic manipulation, including reach and grasp, is essential for deployment into households and workspaces involving diverse and evolving tasks. Recent advances propose using large pre-trained models, such as Large Language Models and object detectors, to boost robotic perception in reinforcement learning. These models, trained on large datasets via self-supervised learning, can p…
▽ More
General-purpose robotic manipulation, including reach and grasp, is essential for deployment into households and workspaces involving diverse and evolving tasks. Recent advances propose using large pre-trained models, such as Large Language Models and object detectors, to boost robotic perception in reinforcement learning. These models, trained on large datasets via self-supervised learning, can process text prompts and identify diverse objects in scenes, an invaluable skill in RL where learning object interaction is resource-intensive. This study demonstrates how to integrate such models into Goal-Conditioned Reinforcement Learning to enable general and versatile robotic reach and grasp capabilities. We use a pre-trained object detection model to enable the agent to identify the object from a text prompt and generate a mask for goal conditioning. Mask-based goal conditioning provides object-agnostic cues, improving feature sharing and generalization. The effectiveness of the proposed framework is demonstrated in a simulated reach-and-grasp task, where the mask-based goal conditioning consistently maintains a $\sim$90\% success rate in grasping both in and out-of-distribution objects, while also ensuring faster convergence to higher returns.
△ Less
Submitted 14 July, 2025;
originally announced July 2025.
-
Photonics in Flatland: Challenges and Opportunities for Nanophotonics with 2D Semiconductors
Authors:
Ali Azimi,
Julien Barrier,
Angela Barreda,
Thomas Bauer,
Farzaneh Bouzari,
Abel Brokkelkamp,
Francesco Buatier de Mongeot,
Timothy Parsons,
Peter Christianen,
Sonia Conesa-Boj,
Alberto G. Curto,
Suprova Das,
Bernardo Dias,
Itai Epstein,
Zlata Fedorova,
F. Javier García de Abajo,
Ilya Goykhman,
Lara Greten,
Johanna Grönqvist,
Ludovica Guarneri,
Yujie Guo,
Tom Hoekstra,
Xuerong Hu,
Benjamin Laudert,
Jason Lynch
, et al. (23 additional authors not shown)
Abstract:
Two-dimensional (2D) semiconductors are emerging as a versatile platform for nanophotonics, offering unprecedented tunability in optical properties through exciton resonance engineering, van der Waals heterostructuring, and external field control. These materials enable active optical modulation, single-photon emission, quantum photonics, and valleytronic functionalities, paving the way for next-g…
▽ More
Two-dimensional (2D) semiconductors are emerging as a versatile platform for nanophotonics, offering unprecedented tunability in optical properties through exciton resonance engineering, van der Waals heterostructuring, and external field control. These materials enable active optical modulation, single-photon emission, quantum photonics, and valleytronic functionalities, paving the way for next-generation optoelectronic and quantum photonic devices. However, key challenges remain in achieving large-area integration, maintaining excitonic coherence, and optimizing amplitude-phase modulation for efficient light manipulation. Advances in fabrication, strain engineering, and computational modelling will be crucial to overcoming these limitations. This perspective highlights recent progress in 2D semiconductor-based nanophotonics, emphasizing opportunities for scalable integration into photonics.
△ Less
Submitted 30 June, 2025;
originally announced July 2025.
-
Deep Policy Gradient Methods Without Batch Updates, Target Networks, or Replay Buffers
Authors:
Gautham Vasan,
Mohamed Elsayed,
Alireza Azimi,
Jiamin He,
Fahim Shariar,
Colin Bellinger,
Martha White,
A. Rupam Mahmood
Abstract:
Modern deep policy gradient methods achieve effective performance on simulated robotic tasks, but they all require large replay buffers or expensive batch updates, or both, making them incompatible for real systems with resource-limited computers. We show that these methods fail catastrophically when limited to small replay buffers or during incremental learning, where updates only use the most re…
▽ More
Modern deep policy gradient methods achieve effective performance on simulated robotic tasks, but they all require large replay buffers or expensive batch updates, or both, making them incompatible for real systems with resource-limited computers. We show that these methods fail catastrophically when limited to small replay buffers or during incremental learning, where updates only use the most recent sample without batch updates or a replay buffer. We propose a novel incremental deep policy gradient method -- Action Value Gradient (AVG) and a set of normalization and scaling techniques to address the challenges of instability in incremental learning. On robotic simulation benchmarks, we show that AVG is the only incremental method that learns effectively, often achieving final performance comparable to batch policy gradient methods. This advancement enabled us to show for the first time effective deep reinforcement learning with real robots using only incremental updates, employing a robotic manipulator and a mobile robot.
△ Less
Submitted 21 May, 2025; v1 submitted 22 November, 2024;
originally announced November 2024.
-
zGAN: An Outlier-focused Generative Adversarial Network For Realistic Synthetic Data Generation
Authors:
Azizjon Azimi,
Bonu Boboeva,
Ilyas Varshavskiy,
Shuhrat Khalilbekov,
Akhlitdin Nizamitdinov,
Najima Noyoftova,
Sergey Shulgin
Abstract:
The phenomenon of "black swans" has posed a fundamental challenge to performance of classical machine learning models. The perceived rise in frequency of outlier conditions, especially in post-pandemic environment, has necessitated exploration of synthetic data as a complement to real data in model training. This article provides a general overview and experimental investigation of the zGAN model…
▽ More
The phenomenon of "black swans" has posed a fundamental challenge to performance of classical machine learning models. The perceived rise in frequency of outlier conditions, especially in post-pandemic environment, has necessitated exploration of synthetic data as a complement to real data in model training. This article provides a general overview and experimental investigation of the zGAN model architecture developed for the purpose of generating synthetic tabular data with outlier characteristics. The model is put to test in binary classification environments and shows promising results on realistic synthetic data generation, as well as uplift capabilities vis-à-vis model performance. A distinctive feature of zGAN is its enhanced correlation capability between features in the generated data, replicating correlations of features in real training data. Furthermore, crucial is the ability of zGAN to generate outliers based on covariance of real data or synthetically generated covariances. This approach to outlier generation enables modeling of complex economic events and augmentation of outliers for tasks such as training predictive models and detecting, processing or removing outliers. Experiments and comparative analyses as part of this study were conducted on both private (credit risk in financial services) and public datasets.
△ Less
Submitted 29 October, 2024; v1 submitted 28 October, 2024;
originally announced October 2024.
-
The maximum four point condition matrix of a tree
Authors:
Ali Azimi,
Rakesh Jana,
Mukesh Kumar Nagar,
Sivaramakrishnan Sivasubramanian
Abstract:
$\newcommand{\Max}{\mathrm{Max4PC}}$ The Four point condition (4PC henceforth) is a well known condition characterising distances in trees $T$. Let $w,x,y,z$ be four vertices in $T$ and let $d_{x,y}$ denote the distance between vertices $x,y$ in $T$. The 4PC condition says that among the three terms $d_{w,x} + d_{y,z}$, $d_{w,y} + d_{x,z}$ and $d_{w,z} + d_{x,y}…
▽ More
$\newcommand{\Max}{\mathrm{Max4PC}}$ The Four point condition (4PC henceforth) is a well known condition characterising distances in trees $T$. Let $w,x,y,z$ be four vertices in $T$ and let $d_{x,y}$ denote the distance between vertices $x,y$ in $T$. The 4PC condition says that among the three terms $d_{w,x} + d_{y,z}$, $d_{w,y} + d_{x,z}$ and $d_{w,z} + d_{x,y}$ the maximum value equals the second maximum value.
We define an $\binom{n}{2} \times \binom{n}{2}$ sized matrix $\Max_T$ from a tree $T$ where the rows and columns are indexed by size-2 subsets. The entry of $\Max_T$ corresponding to the row indexed by $\{w,x\}$ and column $\{y,z\}$ is the maximum value among the three terms $d_{w,x} + d_{y,z}$, $d_{w,y} + d_{x,z}$ and $d_{w,z} + d_{x,y}$. In this work, we determine basic properties of this matrix like rank, give an algorithm that outputs a family of bases, and find the determinant of $\Max_T$ when restricted to our basis. We further determine the inertia and the Smith Normal Form (SNF) of $\Max_T$.
△ Less
Submitted 16 August, 2023;
originally announced August 2023.
-
Explicit formulas for matrices associated to ladder, circular ladder, and Mobius ladder graphs
Authors:
Ali Azimi,
Mohammad Farrokhi Derakhshandeh Ghouchan
Abstract:
We give explicit formulas for resistance distance matrices and Moore-Penrose inverses of incidence and Laplacian matrices of ladder, circular ladder, and Möbius ladder graphs. As a result, we compute the Kirchhoff index of these graphs and give new combinatorial formulas for the number of their spanning trees.
We give explicit formulas for resistance distance matrices and Moore-Penrose inverses of incidence and Laplacian matrices of ladder, circular ladder, and Möbius ladder graphs. As a result, we compute the Kirchhoff index of these graphs and give new combinatorial formulas for the number of their spanning trees.
△ Less
Submitted 18 June, 2023;
originally announced June 2023.
-
Moore-Penrose inverse of incidence matrices
Authors:
Ali Azimi,
R. B. Bapat,
Mohammad Farrokhi Derakhshandeh Ghouchan
Abstract:
We present explicit formulas for Moore-Penrose inverses of some families of set inclusion matrices arising from sets, vector spaces, and designs.
We present explicit formulas for Moore-Penrose inverses of some families of set inclusion matrices arising from sets, vector spaces, and designs.
△ Less
Submitted 22 May, 2023;
originally announced May 2023.
-
Steiner distance matrix of caterpillar graphs
Authors:
Ali Azimi,
R. B. Bapat,
Shivani Goel
Abstract:
For a connected graph $G:=(V,E)$, the Steiner distance $d_G(X)$ among a set of vertices $X$ is the minimum size among all the connected subgraphs of $G$ whose vertex set contains $X$. The $k-$Steiner distance matrix $D_k(G)$ of $G$ is a matrix whose rows and columns are indexed by $k-$subsets of $V$. For $k$-subsets $X_1$ and $X_2$, the $(X_1,X_2)-$entry of $D_k(G)$ is $d_G(X_1 \cup X_2)$. In this…
▽ More
For a connected graph $G:=(V,E)$, the Steiner distance $d_G(X)$ among a set of vertices $X$ is the minimum size among all the connected subgraphs of $G$ whose vertex set contains $X$. The $k-$Steiner distance matrix $D_k(G)$ of $G$ is a matrix whose rows and columns are indexed by $k-$subsets of $V$. For $k$-subsets $X_1$ and $X_2$, the $(X_1,X_2)-$entry of $D_k(G)$ is $d_G(X_1 \cup X_2)$. In this paper, we show that the rank of $2-$Steiner distance matrix of a caterpillar graph on $N$ vertices and with $p$ pendant veritices is $2N-p-1$.
△ Less
Submitted 29 August, 2021;
originally announced August 2021.
-
Factorization graphs of finite groups
Authors:
Mohammad Farrokhi Derakhshandeh Ghouchan,
Ali Azimi
Abstract:
We introduce the factorization graph of a finite group and study its connectedness and forbidden structures. We characterize all finite groups with connected factorization graphs and classify those with connected bipartite factorization graphs. Also, we obtain a classification of all groups with claw-free, $K_{1,4}$-free, and square-free factorization graphs, excluding the non-solvable groups in t…
▽ More
We introduce the factorization graph of a finite group and study its connectedness and forbidden structures. We characterize all finite groups with connected factorization graphs and classify those with connected bipartite factorization graphs. Also, we obtain a classification of all groups with claw-free, $K_{1,4}$-free, and square-free factorization graphs, excluding the non-solvable groups in the latter case.
△ Less
Submitted 28 November, 2019;
originally announced November 2019.
-
Self 2-distance graphs
Authors:
Ali Azimi,
Mohammad Farrokhi Derakhshandeh Ghouchan
Abstract:
All finite simple self $2$-distance graphs with no $4$-cycle, diamond, or triangles with a common vertex are determined. Utilizing these results, it is shown that there is no cubic self $2$-distance graphs.
All finite simple self $2$-distance graphs with no $4$-cycle, diamond, or triangles with a common vertex are determined. Utilizing these results, it is shown that there is no cubic self $2$-distance graphs.
△ Less
Submitted 13 October, 2015;
originally announced October 2015.
-
On cycles in intersection graph of rings
Authors:
Ali Azimi,
Ahmad Erfanian,
Mohammad Farrokhi Derakhshandeh Ghouchan,
Nesa Hoseini
Abstract:
Let $R$ be a commutative ring with non-zero identity. We describe all $C_3$- and $C_4$-free intersection graph of non-trivial ideals of $R$ as well as $C_n$-free intersection graph when $R$ is a reduced ring. Also, we shall describe all complete, regular and $n$-claw-free intersection graphs. Finally, we shall prove that almost all Artin rings $R$ have Hamiltonian intersection graphs. We show that…
▽ More
Let $R$ be a commutative ring with non-zero identity. We describe all $C_3$- and $C_4$-free intersection graph of non-trivial ideals of $R$ as well as $C_n$-free intersection graph when $R$ is a reduced ring. Also, we shall describe all complete, regular and $n$-claw-free intersection graphs. Finally, we shall prove that almost all Artin rings $R$ have Hamiltonian intersection graphs. We show that such graphs are indeed pancyclic.
△ Less
Submitted 26 January, 2014;
originally announced January 2014.
-
Cycles and paths in Jacobson graphs
Authors:
Ali Azimi,
Mohammad Farrokhi Derakhshandeh Ghouchan
Abstract:
All finite Jacobson graphs with a Hamiltonian cycle or path, or Eulerian tour or trail are determined, and it is shown that a finite Jacobson graph is Hamiltonian if and only if it is pancyclic. Also, the length of the longest induced cycles and paths in finite Jacobson graphs are obtained.
All finite Jacobson graphs with a Hamiltonian cycle or path, or Eulerian tour or trail are determined, and it is shown that a finite Jacobson graph is Hamiltonian if and only if it is pancyclic. Also, the length of the longest induced cycles and paths in finite Jacobson graphs are obtained.
△ Less
Submitted 26 January, 2014;
originally announced January 2014.
-
Isomorphisms between Jacobson graphs
Authors:
Ali Azimi,
Ahmad Erfanian,
Mohammad Farrokhi Derakhshandeh Ghouchan
Abstract:
Let $R$ be a commutative ring with a non-zero identity and $\mathfrak{J}_R$ be its Jacobson graph. We show that if $R$ and $R'$ are finite commutative rings, then $\mathfrak{J}_R\cong\mathfrak{J}_{R'}$ if and only if $|J(R)|=|J(R')|$ and $R/J(R)\cong R'/J(R')$. Also, for a Jacobson graph $\mathfrak{J}_R$, we obtain the structure of group $\mathrm{Aut}(\mathfrak{J}_R)$ of all automorphisms of…
▽ More
Let $R$ be a commutative ring with a non-zero identity and $\mathfrak{J}_R$ be its Jacobson graph. We show that if $R$ and $R'$ are finite commutative rings, then $\mathfrak{J}_R\cong\mathfrak{J}_{R'}$ if and only if $|J(R)|=|J(R')|$ and $R/J(R)\cong R'/J(R')$. Also, for a Jacobson graph $\mathfrak{J}_R$, we obtain the structure of group $\mathrm{Aut}(\mathfrak{J}_R)$ of all automorphisms of $\mathfrak{J}_R$ and prove that under some conditions two semi-simple rings $R$ and $R'$ are isomorphic if and only if $\mathrm{Aut}(\mathfrak{J}_R)\cong\mathrm{Aut}(\mathfrak{J}_{R'})$.
△ Less
Submitted 25 January, 2014;
originally announced January 2014.
-
Applying the Apriori algorithm for investigating the relationships between demographic characteristics of Iranian top 100 enterprises and the strcture of their commercial website
Authors:
Ali Azimi,
Azar Kaffashpour
Abstract:
This study was conducted with the main aim to investigate the relationships between demographic characteristics of companies and the facilities required for their commercial websites. The research samples are the top 100 Iranian companies as ranked by the Iranian Industrial Management Institute; the method applied is datamining, using Association Rules throught the Apriori algorithms. To collect t…
▽ More
This study was conducted with the main aim to investigate the relationships between demographic characteristics of companies and the facilities required for their commercial websites. The research samples are the top 100 Iranian companies as ranked by the Iranian Industrial Management Institute; the method applied is datamining, using Association Rules throught the Apriori algorithms. To collect the data, an aithor-modified check list has been utilized, coverig the three areas of faclities within commercial websites, i.e. fundamental, information-providing, and service-delivering facilities. having extracted the association rules between the mentioned two sets of variables, 68 rules with a confidence rate of 90% and above were obtained, and based on their significance were classified into two groups of must-have and should-have requirements; a recommended package of facilities is hitherto offered to other companies which intend to enter e-commerce through their commerical websites with regards to each company's unique demographic characteristics.
△ Less
Submitted 7 December, 2013;
originally announced December 2013.
